def makeFloHtml(out_file_html,img_files,flo_files,height=200,width=200):
img_paths=[];
captions=[];
for img_file,flo_file in zip(img_files,flo_files):
img_path=[];
img_path.append(util.getRelPath(img_file,'/disk2'));
img_path.append(util.getRelPath(flo_file,'/disk2'));
img_paths.append(img_path);
captions.append(['img','flo']);
visualize.writeHTML(out_file_html,img_paths,captions,height,width);
def makeFloVizHTML(out_file_html,img_paths,dir_flo_viz):
# out_file_html=os.path.join(out_dir,'flo_viz.html');
img_paths_html=[];
captions_html=[];
for img_path,img_file_name in zip(img_paths,util.getFileNames(img_paths,ext=False)):
out_file_flo_viz=os.path.join(dir_flo_viz,img_file_name+'.png');
if img_path.startswith('/disk2'):
img_path='/disk3'+img_path;
img_paths_curr=[util.getRelPath(img_path,'/disk3'),util.getRelPath(out_file_flo_viz,'/disk3')];
img_paths_html.append(img_paths_curr);
captions_html.append([img_file_name,'flo']);
visualize.writeHTML(out_file_html,img_paths_html,captions_html);
def script_testOnYoutube():
val_file='/disk2/mayExperiments/finetuning_youtube_hmdb_llr/val_eq.txt'
out_dir='/disk2/mayExperiments/eval_ucf_finetune';
clusters_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/clusters.mat';
gpu=0;
util.mkdir(out_dir);
# model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/final.caffemodel';
# out_dir_model=os.path.join(out_dir,'original_model');
model_file='/disk2/mayExperiments/ft_youtube_hmdb_ucfClusters/OptFlow_youtube_hmdb__iter_55000.caffemodel';
out_dir_model=os.path.join(out_dir,'ft_ucf_model');
util.mkdir(out_dir_model);
out_dir_flo=os.path.join(out_dir_model,'flo');
out_dir_flo_viz=os.path.join(out_dir_model,'flo_viz');
util.mkdir(out_dir_flo);util.mkdir(out_dir_flo_viz)
num_to_pick=20;
img_paths=util.readLinesFromFile(val_file);
img_paths=[img_path[:img_path.index(' ')] for img_path in img_paths];
class_names=[file_curr[:file_curr.index('_')] for file_curr in util.getFileNames(img_paths)];
classes=list(set(class_names));
class_names=np.array(class_names);
img_paths_test=[];
for class_curr in classes:
idx_rel=np.where(class_names==class_curr)[0];
idx_rel=idx_rel[:num_to_pick];
img_paths_test.extend([img_paths[idx_curr] for idx_curr in idx_rel]);
# po.script_saveFlosAndViz(img_paths_test,out_dir_flo,out_dir_flo_viz,gpu,model_file,clusters_file);
out_file_html=os.path.join(out_dir,'model_comparison.html');
out_dirs_flo_viz=[os.path.join(out_dir,'original_model','flo_viz'),os.path.join(out_dir,'ft_ucf_model','flo_viz')];
out_dirs_flo_viz_captions=['original_model','ft_ucf_model'];
img_paths_html=[];
captions_html=[];
img_names=util.getFileNames(img_paths_test,ext=False);
for img_path_test,img_name in zip(img_paths_test,img_names):
row_curr=[];
row_curr.append(util.getRelPath(img_path_test));
for out_dir_curr in out_dirs_flo_viz:
file_curr=os.path.join(out_dir_curr,img_name+'.png');
row_curr.append(util.getRelPath(file_curr));
captions_curr=[img_name]+out_dirs_flo_viz_captions;
img_paths_html.append(row_curr)
captions_html.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths_html,captions_html);
def writeHTMLForDifferentFolders(out_file_html,
folders,
captions,
img_names,
rel_path_replace=None,
height=200,
width=200):
if rel_path_replace is None:
string_curr = folders[0]
rel_path_replace = string_curr[:string_curr[1:].index('/') + 1]
# print rel_path_replace
img_paths = []
captions_html = []
for img_name in img_names:
captions_row = []
img_paths_row = []
for caption_curr, folder in zip(captions, folders):
img_paths_row.append(
util.getRelPath(os.path.join(folder, img_name),
rel_path_replace))
captions_row.append(caption_curr)
img_paths.append(img_paths_row)
captions_html.append(captions_row)
writeHTML(out_file_html,
img_paths,
captions_html,
height=height,
width=width)
def script_writeHTMLStitchedFlos_wDirs(img_paths,out_file_html,viz_dirs):
img_paths_html=[];
captions=[];
for img_path in img_paths:
img_name=img_path[img_path.rindex('/')+1:img_path.rindex('.')];
img_paths_html_curr=[util.getRelPath(img_path)];
captions_curr=['im']
for viz_dir in viz_dirs:
print viz_dir,img_path
# img_path_curr=[os.path.join(viz_dir,file_curr) for file_curr in os.listdir(viz_dir) if file_curr.startswith(img_name)][0];
img_path_curr=os.path.join(viz_dir,img_name+'.png');
img_paths_html_curr.append(util.getRelPath(img_path_curr));
captions_curr.append(viz_dir[viz_dir.rindex('/')+1:]);
img_paths_html.append(img_paths_html_curr);
captions.append(captions_curr)
visualize.writeHTML(out_file_html,img_paths_html,captions);
def script_writeHTMLStitchedFlos(out_file_html,out_file,out_dir,grid_sizes=[1,2,4,8],grid_dir_pre='grid_flo_viz_'):
img_paths=util.readLinesFromFile(out_file);
viz_dirs=[os.path.join(out_dir,grid_dir_pre+str(num)) for num in grid_sizes];
img_paths_html=[];
captions=[];
for img_path in img_paths:
img_name=img_path[img_path.rindex('/')+1:img_path.rindex('.')];
img_paths_html_curr=[util.getRelPath(img_path)];
captions_curr=['im']
for viz_dir in viz_dirs:
print viz_dir,img_path
img_path_curr=[os.path.join(viz_dir,file_curr) for file_curr in os.listdir(viz_dir) if file_curr.startswith(img_name)][0];
img_paths_html_curr.append(util.getRelPath(img_path_curr));
captions_curr.append(viz_dir[viz_dir.rindex('/')+1:]);
img_paths_html.append(img_paths_html_curr);
captions.append(captions_curr)
visualize.writeHTML(out_file_html,img_paths_html,captions);
def script_doEverything(path_to_npy,path_to_im,ext,lim,out_file,out_dir_scratch,window,step_size,thresh,scale_info,scale_all, scale_images,lim_cases,gpu,model_file,clusters_file,train_val_file=None,overwrite=False):
for scale in scale_all:
# script_saveImCrops(path_to_npy,path_to_im,ext,lim,out_file,out_dir_scratch,window,step_size,thresh,scale_info,scale,scale_images,lim_cases)
for scale_image in scale_images:
dir_scale=os.path.join(out_dir_scratch,scale+'_'+str(scale_image));
scale_info=pickle.load(open(out_file,'rb'));
img_dirs=[os.path.join(dir_scale,im_curr_info[0]) for im_curr_info in scale_info[scale][:lim_cases]]
for img_dir in img_dirs:
img_paths=util.getFilesInFolder(img_dir,ext='.png');
if len(img_paths)==0:
'CONTINUING'
continue;
img_paths=[img_path for img_path in img_paths if not img_path.endswith('onImg.png')];
out_dir_flo=img_dir+'_pred_flo';
out_dir_flo_viz=img_dir+'_pred_flo_viz';
util.mkdir(out_dir_flo);
util.mkdir(out_dir_flo_viz);
po.script_saveFlosAndViz(img_paths,out_dir_flo,out_dir_flo_viz,gpu,model_file,clusters_file,train_val_file=train_val_file,overwrite=overwrite)
img_names=util.getFileNames(img_paths,ext=False);
out_dir_flo=img_dir+'_pred_flo';
out_dir_flo_viz=img_dir+'_pred_flo_viz';
out_file_html=img_dir+'.html';
img_paths_html=[];
captions_all=[];
for img_name in img_names:
row_curr=[];
row_curr.append(util.getRelPath(os.path.join(img_dir,img_name+'_onImg.png')));
row_curr.append(util.getRelPath(os.path.join(img_dir,img_name+'.png')));
row_curr.append(util.getRelPath(os.path.join(out_dir_flo_viz,img_name+'.png')));
captions=['','','']
img_paths_html.append(row_curr);
captions_all.append(captions);
visualize.writeHTML(out_file_html,img_paths_html,captions_all);
def script_writeFloVizHTML(out_file_html,out_dir_viz,flo_files,im_files,tif_files,clusters,tifAsPng=False):
img_paths=[];
captions=[];
# idx=0;
for flo_file,im_file,tif_file in zip(flo_files,im_files,tif_files):
# print idx;
# print tif_file
assert os.path.exists(tif_file);
assert os.path.exists(im_file);
# print tif_file
# if not os.path.exists(tif_file) or not os.path.exists(im_file) :
# continue;
file_name=util.getFileNames([flo_file],ext=False)[0];
out_file_pre=os.path.join(out_dir_viz,file_name);
out_file_flo_viz=out_file_pre+'_flo.png';
out_files_tif=[out_file_pre+'_tifim_x.png',out_file_pre+'_tifim_y.png',out_file_pre+'_tifflo.png'];
if not os.path.exists(out_file_flo_viz):
flo=util.readFlowFile(flo_file);
po.saveFloFileViz(flo_file,out_file_flo_viz);
for idx,out_file_tif_viz in enumerate(out_files_tif):
tif=scipy.misc.imread(tif_file)[:,:,:2];
if idx==0 and not os.path.exists(out_file_tif_viz):
tif_flo=replaceClusterIdWithFlow(tif,clusters);
po.saveMatFloViz(tif_flo,out_file_tif_viz);
if not os.path.exists(out_file_tif_viz) and idx==1:
tif_x=np.array(tif[:,:,0]*(255.0/clusters.shape[0]),dtype=int);
tif_x=np.dstack((tif_x,tif_x,tif_x));
scipy.misc.imsave(out_file_tif_viz,tif_x);
if not os.path.exists(out_file_tif_viz) and idx==2:
tif_x=np.array(tif[:,:,1]*(255.0/clusters.shape[0]),dtype=int);
tif_x=np.dstack((tif_x,tif_x,tif_x));
scipy.misc.imsave(out_file_tif_viz,tif_x);
img_paths_curr=[im_file,out_file_flo_viz]+out_files_tif;
im_name=util.getFileNames([im_file],ext=False)[0];
captions_curr=[im_name,'flo_viz']+['tif_flo_viz']*len(out_files_tif)
# if tifAsPng:
# img_paths_curr.append(out_file_tif_viz.replace('_x.png','_y.png'));
# captions_curr.append('tif_flo_viz');
img_paths_curr=[util.getRelPath(file_curr) for file_curr in img_paths_curr];
img_paths.append(img_paths_curr);
captions.append(captions_curr);
# idx=idx+1;
visualize.writeHTML(out_file_html,img_paths,captions)
def comparativeLossViz(img_dirs,
file_post,
loss_post,
range_batches,
range_images,
out_file_html,
dir_server,
img_caption_pre=None):
img_files_all = []
captions_all = []
if img_caption_pre is not None:
assert len(img_caption_pre) == len(img_dirs)
for batch_num in range_batches:
# range(1,num_batches+1):
for im_num in range_images:
for idx_img_dir, img_dir in enumerate(img_dirs):
loss_all = np.load(
os.path.join(img_dir,
str(batch_num) + loss_post))
if im_num > loss_all.shape[0]:
continue
loss_curr = loss_all[im_num - 1, 0]
loss_str = "{:10.4f}".format(loss_curr)
files_curr = [
os.path.join(
img_dir,
str(batch_num) + '_' + str(im_num) + file_post_curr)
for file_post_curr in file_post
]
files_curr = [
util.getRelPath(file_curr, dir_server)
for file_curr in files_curr
]
captions_curr = [
os.path.split(file_curr)[1] + ' ' + loss_str
for file_curr in files_curr
]
if img_caption_pre is not None:
captions_curr = [
img_caption_pre[idx_img_dir] + ' ' + caption_curr
for caption_curr in captions_curr
]
img_files_all.append(files_curr)
captions_all.append(captions_curr)
visualize.writeHTML(out_file_html, img_files_all, captions_all, 224, 224)
def makeImTifViz(img_paths_all,tif_paths_all,out_file_html,out_dir_tif,num_clusters=40,disk_path='/disk2'): out_files_tif_x=[os.path.join(out_dir_tif,img_name+'_x.png') for img_name in util.getFileNames(tif_paths_all,ext='False')]; out_files_tif_y=[os.path.join(out_dir_tif,img_name+'_y.png') for img_name in util.getFileNames(tif_paths_all,ext='False')]; for tif_path,out_file_x,out_file_y in zip(tif_paths_all,out_files_tif_x,out_files_tif_y): tif=scipy.misc.imread(tif_path); # print np.min(tif[:,:,:2]),np.max(tif[:,:,:2]) assert np.min(tif[:,:,:2])>0 and np.max(tif[:,:,:2])<num_clusters+1; saveTifGray(tif,out_file_x,out_file_y,num_clusters) # out_file_html=out_dir_tif+'.html'; img_paths_html=[[util.getRelPath(img_curr,disk_path) for img_curr in img_list] for img_list in zip(img_paths_all,out_files_tif_x,out_files_tif_y)]; # captions_html=[[util.getFileNames([img_curr],ext=False)[0] for img_curr in img_list] for img_list in zip(img_paths_all,out_files_tif_x,out_files_tif_y)]; captions_html=[['Image','Tif_x','Tif_y']]*len(img_paths_html); visualize.writeHTML(out_file_html,img_paths_html,captions_html);
def makeClusterHTML(out_file_html, labels, num_cols, size_im, dir_server):
ims = []
captions = []
start_idx = 0
while start_idx < len(labels):
row_curr = []
caption_curr = []
if start_idx + num_cols > len(labels):
num_cols_real = len(labels) - start_idx
else:
num_cols_real = num_cols
for col_no in range(num_cols_real):
idx_curr = start_idx + col_no
label_curr = labels[idx_curr]
row_curr.append(util.getRelPath(label_curr, dir_server))
caption_curr.append('')
ims.append(row_curr)
captions.append(caption_curr)
start_idx = start_idx + num_cols_real
visualize.writeHTML(out_file_html, ims, captions, size_im, size_im)
print out_file_html.replace(dir_server,
'http://vision1.idav.ucdavis.edu:1000')
def compareMagInfoLayerTime(dirs,out_dirs,model_num,num_iters,file_pre_grad,file_pre_weight,out_file_html):
img_paths=[];
captions=[];
for dir_curr,out_dir_curr in zip(dirs,out_dirs):
dict_for_plotting={'grads_mag':OrderedDict(),'weights_mag':OrderedDict(),'ratios':OrderedDict()};
dict_for_plotting=OrderedDict(dict_for_plotting.items());
for model_num_curr in model_num:
print model_num_curr;
file_pairs=[(os.path.join(dir_curr,model_num_curr,file_pre_grad+str(iter_curr)+'.npy'),
os.path.join(dir_curr,model_num_curr,file_pre_weight+str(iter_curr)+'.npy')) for iter_curr in num_iters];
means,_=getMagInfo(file_pairs,alt=True);
dict_for_plotting['grads_mag'][model_num_curr]=means[0];
dict_for_plotting['weights_mag'][model_num_curr]=means[1];
dict_for_plotting['ratios'][model_num_curr]=means[2];
img_paths_curr=[];
captions_curr=[];
for key_curr in dict_for_plotting.keys():
out_file_curr=os.path.join(out_dir_curr,key_curr+'.png');
data=dict_for_plotting[key_curr];
xAndYs=data.values();
legend_entries=data.keys();
xAndYs=[(range(len(x_curr)),x_curr) for x_curr in xAndYs];
visualize.plotSimple(xAndYs,out_file_curr,title=key_curr,xlabel='layer',ylabel='magnitude',legend_entries=legend_entries,outside=True);
print out_file_curr.replace('/disk3','vision3.cs.ucdavis.edu:1001');
img_paths_curr.append(util.getRelPath(out_file_curr,'/disk3'));
# print dir_curr.split('/');
captions_curr.append(dir_curr.split('/')[-2]+' '+dir_curr.split('/')[-1]+' '+key_curr);
img_paths.append(img_paths_curr);
captions.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths,captions,height=500,width=500);
print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
def script_visualizeLossesFromExperiment():
# out_dir='/disk3/maheen_data/ft_youtube_40_noFix_alexnet';
out_dir='/disk3/maheen_data/debug_networks/noFixCopyByLayerAlexNet';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/models/bvlc_alexnet/bvlc_alexnet.caffemodel';
layers=['conv1','conv2','conv3','conv4','conv5','fc6'];
# ,'fc7'];
layers_str=[];
for idx in range(len(layers)):
# if idx==0:
# fix_layers=layers[0];
# layer_str=str(fix_layers);
# else:
fix_layers=layers[:idx+1];
layer_str='_'.join(fix_layers);
layers_str.append(layer_str);
log_files=[os.path.join(out_dir,'log_'+layer_str+'.log') for layer_str in layers_str];
str_match=' solver.cpp:209] Iteration ';
xAndYs=[svl.getIterationsAndLosses(log_file,str_match) for log_file in log_files];
out_files=[];
for layer_str,log_file in zip(layers_str,log_files):
xAndY=svl.getIterationsAndLosses(log_file,str_match);
print xAndY
out_file=os.path.join(out_dir,'loss_'+layer_str+'.png');
visualize.plotSimple([xAndY],out_file,title=layer_str);
out_files.append(out_file);
out_file_html=os.path.join(out_dir,'losses_all.html');
img_paths=[[util.getRelPath(out_file,'/disk3')] for out_file in out_files];
captions=[['']]*len(out_files);
print img_paths
print captions
visualize.writeHTML(out_file_html,img_paths,captions,height=300,width=300);
def script_writeHTMLForOverlap(path_to_npy,path_to_im,ext,lim,out_file,out_dir_scratch,window,step_size,thresh,scale_info,scale_all, scale_images,lim_cases,gpu,model_file,clusters_file):
img_dirs_all=[];
for scale in scale_all:
for scale_image in scale_images:
dir_scale=os.path.join(out_dir_scratch,scale+'_'+str(scale_image));
scale_info=pickle.load(open(out_file,'rb'));
img_dirs=[os.path.join(dir_scale,im_curr_info[0]) for im_curr_info in scale_info[scale][:lim_cases]]
img_dirs_all=img_dirs_all+img_dirs;
# record_files_all=[];
img_files_record={};
for img_dir in img_dirs_all:
record_files=[os.path.join(img_dir,file_curr) for file_curr in util.getFilesInFolder(img_dir,'.p')];
for record_file in record_files:
record=pickle.load(open(record_file,'rb'));
if len(record[0])==0:
continue;
img_name=img_dir[img_dir.rindex('/')+1:];
rel_files=[];
# print len(record[0])
# print record_file
img_name_ac=record_file[record_file.rindex('/')+1:record_file.rindex('_')];
for idx_curr in range(len(record[0])):
rel_file_curr=[];
rel_file_curr.append(os.path.join(img_dir,img_name_ac+'_'+str(idx_curr)+'_onImg.png'));
rel_file_curr.append(os.path.join(img_dir,img_name_ac+'_'+str(idx_curr)+'.png'));
rel_file_curr.append(os.path.join(img_dir+'_pred_flo_viz',img_name_ac+'_'+str(idx_curr)+'.png'));
rel_file_curr.append(record[1][idx_curr]);
rel_files.append(rel_file_curr);
if img_name_ac in img_files_record:
img_files_record[img_name_ac].extend(rel_files);
else:
img_files_record[img_name_ac]=rel_files;
print len(img_files_record);
# print img_files_record[img_files_record.keys()[0]];
out_file_html=os.path.join(out_dir_scratch,'visualize.html');
img_paths_html=[];
captions_html=[];
for img_name in img_files_record.keys():
img_paths_row=[];
captions_row=[];
rec=img_files_record[img_name];
rec_np=np.array(rec);
print rec_np.shape
sort_idx=np.argsort(rec_np[:,-1])[::-1];
for idx_curr in sort_idx[::5]:
img_paths_row.extend(rec[idx_curr][:-1]);
captions_row.extend([str(rec[idx_curr][-1]),str(rec[idx_curr][-1]),str(rec[idx_curr][-1])]);
img_paths_row=[util.getRelPath(path_curr) for path_curr in img_paths_row];
img_paths_html.append(img_paths_row);
captions_html.append(captions_row);
visualize.writeHTML(out_file_html,img_paths_html,captions_html);
def main():
# dirs=['/disk3/maheen_data/headC_160_withFlow_human_xavier_unit_floStumpPretrained_fullTraining/gradient_checks'];
# # dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct_16']];
# range_flow=range(1);
# out_dirs=[os.path.join(dir_curr,'plots') for dir_curr in dirs];
# [util.mkdir(out_dir_curr) for out_dir_curr in out_dirs];
# model_num=range(5000,45000,5000);
# model_num.append(45000);
# print model_num
# model_num=[str(model_num_curr) for model_num_curr in model_num]
# num_iters=range(1,5);
# # num_iters=range(2,3);
# file_pre_weight='weight_mag_n_';
# file_pre_grad='grad_mag_n_';
# out_file_html=os.path.join(dirs[0],'comparison_grads_weights_ratios_n.html');
# compareMagInfoLayerTime(dirs,out_dirs,model_num,num_iters,file_pre_grad,file_pre_weight,out_file_html)
# out_file_html=os.path.join(dirs[0],'comparison_grads_seg_no_seg_16.html');
# layer_range=[26,27,28,31];
# num_iters=range(1,5,2);
# img_paths_seg_flow=compareGradInfoLayer([dirs[i] for i in range_flow],[out_dirs[i] for i in range_flow],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
# layer_range=range(26,32);
# num_iters=range(2,5,2);
# img_paths_score_flow=compareGradInfoLayer([dirs[i] for i in range_flow],[out_dirs[i] for i in range_flow],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
# img_paths=[img_paths_seg_flow,img_paths_score_flow];
# img_paths=[[util.getRelPath(path_curr,'/disk3') for path_curr in list_curr] for list_curr in img_paths];
# captions=[];
# for list_curr in img_paths:
# captions_curr=[];
# for path in list_curr:
# path_split=path.split('/');
# caption=path_split[-4]+' '+path_split[-3];
# captions_curr.append(caption);
# print caption
# captions.append(captions_curr);
# visualize.writeHTML(out_file_html,img_paths,captions,height=500,width=500);
# print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
# return
# model_dir='/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_floStumpPretrained_fullTraining/intermediate'
# out_dir_meta='/disk3/maheen_data/headC_160_withFlow_human_xavier_unit_floStumpPretrained_fullTraining'
# out_dir=os.path.join(out_dir_meta,'gradient_checks')
# util.mkdir(out_dir);
# params=[(model_dir,out_dir,'40')];
# out_file_commands_pre=os.path.join(out_dir_meta,'debug_commands_');
# path_to_train_file='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_withFlow_debug.th'
# model_num=range(5000,45000,5000);
# model_num.append(45000);
# print model_num
# # return
# writeCommandsForTrainDebug(params,path_to_train_file,out_file_commands_pre,model_num)
# return
dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug'];
dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct']];
range_flow=range(1);
# range_noflow=range(2,len(dirs));
# dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug','/disk3/maheen_data/headC_160/noFlow_human_debug'];
# dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct_16','incorrect']];
# range_flow=range(2);
# range_noflow=range(2,len(dirs));
out_dirs=[os.path.join(dir_curr,'plots') for dir_curr in dirs];
[util.mkdir(out_dir_curr) for out_dir_curr in out_dirs];
model_num=range(5000,100000,20000);
model_num.append(100000);
# model_num=range(2000,32000,6000);
# model_num.append(32000);
model_num=[str(model_num_curr) for model_num_curr in model_num]
print model_num
# num_iters=range(1,21);
num_iters=range(2,3);
file_pre_weight='weight_mag_n_';
file_pre_grad='grad_mag_n_';
out_file_html=os.path.join('/disk3/maheen_data/headC_160/withFlow_human_debug','comparison_grads_weights_ratios.html');
compareMagInfoLayerTime(dirs,out_dirs,model_num,num_iters,file_pre_grad,file_pre_weight,out_file_html)
# compareMagInfoLayerTime(dirs,out_dirs,model_num,num_iters,file_pre_grad,file_pre_weight,out_file_html)
out_file_html=os.path.join('/disk3/maheen_data/headC_160/withFlow_human_debug','comparison_grads_seg_no_seg.html');
layer_range=[26,27,28,31];
# layer_range=[27,28];
num_iters=range(3,21,2);
img_paths_seg_flow=compareGradInfoLayer([dirs[i] for i in range_flow],[out_dirs[i] for i in range_flow],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
layer_range=range(26,32);
num_iters=range(2,21,2);
img_paths_score_flow=compareGradInfoLayer([dirs[i] for i in range_flow],[out_dirs[i] for i in range_flow],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
# layer_range=[13,14,17]
# num_iters=range(1,21,2);
# img_paths_seg_noflow=compareGradInfoLayer([dirs[i] for i in range_noflow],[out_dirs[i] for i in range_noflow],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
# layer_range=range(13,18);
# num_iters=range(2,21,2);
# img_paths_score_noflow=compareGradInfoLayer([dirs[i] for i in range_noflow],[out_dirs[i] for i in range_noflow],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
# layer_range=[26,27,28,31];
# num_iters=range(1,21,2);
# img_paths_seg_flow=compareGradInfoLayer(dirs[1:],out_dirs[1:],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
# layer_range=range(26,32);
# num_iters=range(2,21,2);
# num_iters=range(2,7,2);
# img_paths_score_flow=compareGradInfoLayer(dirs[1:],out_dirs[1:],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
# img_paths=[img_paths_seg_flow,img_paths_score_flow,img_paths_seg_noflow,img_paths_score_noflow];
img_paths=[img_paths_seg_flow,img_paths_score_flow];
img_paths=[[util.getRelPath(path_curr,'/disk3') for path_curr in list_curr] for list_curr in img_paths];
captions=[];
for list_curr in img_paths:
captions_curr=[];
for path in list_curr:
path_split=path.split('/');
caption=path_split[-4]+' '+path_split[-3];
captions_curr.append(caption);
print caption
captions.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths,captions,height=500,width=500);
print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
return
params=[('/disk3/maheen_data/headC_160/withFlow_xavier_16_score/intermediate','/disk3/maheen_data/headC_160/withFlow_human_debug/correct_16','40')];
out_file_commands_pre='/disk3/maheen_data/headC_160/withFlow_human_debug/debug_commands_16_';
path_to_train_file='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_withFlow_debug.th'
# params=[('/disk3/maheen_data/headC_160/noFlow_gaussian_human_softmax/intermediate_res','/disk3/maheen_data/headC_160/noFlow_human_debug/correct','40'),
# ('/disk3/maheen_data/headC_160/noFlow_gaussian_human/intermediate','/disk3/maheen_data/headC_160/noFlow_human_debug/incorrect','56')];
# [util.mkdir(params_curr[1]) for params_curr in params];
# out_file_commands_pre='/disk3/maheen_data/headC_160/noFlow_human_debug/debug_commands_';
# path_to_train_file='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_noFlow_debug.th'
model_num=range(2000,32000,6000);
model_num.append(32000);
print model_num
return
writeCommandsForTrainDebug(params,path_to_train_file,out_file_commands_pre,model_num)
# maheen_data/headC_160/withFlow_xavier_16_score/intermediate/
return
dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug','/disk3/maheen_data/headC_160/noFlow_human_debug'];
# out_file_html=os.path.join(dirs[0],'comparison_dloss_seg_score.html');
dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct']];
out_dirs=[os.path.join(dir_curr,'plots') for dir_curr in dirs];
file_seg = 'loss_seg.npy';
file_score = 'loss_score.npy';
num_iters=range(2,21);
model_num=range(5000,100000,20000);
model_num.append(100000);
model_num=[str(model_num_curr) for model_num_curr in model_num]
img_paths=[];
captions=[];
for dir_curr,out_dir_curr in zip(dirs,out_dirs):
dict_for_plotting={'loss_seg_all':OrderedDict(),'loss_score_all':OrderedDict(),'loss_ratio_all':OrderedDict()};
for model_num_curr in model_num:
file_curr_seg=os.path.join(dir_curr,model_num_curr,file_seg);
file_curr_score=os.path.join(dir_curr,model_num_curr,file_score);
score_all=np.load(file_curr_score);
score_all=score_all[[0]+range(1,len(score_all),2)];
score_all=score_all*32
seg_all=np.load(file_curr_seg);
seg_all=seg_all[range(0,len(seg_all),2)];
ratios=seg_all/score_all;
print dir_curr,model_num_curr
print np.mean(score_all),np.mean(seg_all),np.mean(ratios);
# break;
# break;
# if num_iter_curr==2:
# score_all.append(np.load(file_curr_score));
# seg_all.append(np.load(file_curr_seg));
# elif num_iter_curr%2==0:
# score_all.append(np.load(file_curr_score));
# else:
# seg_all.append(np.load(file_curr_seg));
# seg_curr=np.load(file_curr_seg);
# seg_curr=np.unique(np.ravel(seg_curr));
# score_curr=list(np.load(file_curr_score));
# score_curr=list(np.unique(np.ravel(score_curr)));
# seg_all.extend(seg_curr);
# score_all.extend(score_curr);
# seg_all=list(set(seg_all));
# score_all=list(set(score_all));
# seg_all.sort();
# score_all.sort();
# dict_for_plotting['seg_all'][model_num_curr]=seg_all;
# dict_for_plotting['score_all'][model_num_curr]=score_all;
# img_paths_curr=[];
# captions_curr=[];
# for key_curr in dict_for_plotting.keys():
# out_file_curr=os.path.join(out_dir_curr,key_curr+'.png');
# data=dict_for_plotting[key_curr];
# xAndYs=data.values();
# legend_entries=data.keys();
# xAndYs=[(range(len(x_curr)),x_curr) for x_curr in xAndYs];
# visualize.plotSimple(xAndYs,out_file_curr,title=key_curr,xlabel='sorted idx',ylabel='values',legend_entries=legend_entries,outside=True);
# print out_file_curr.replace('/disk3','vision3.cs.ucdavis.edu:1001');
# img_paths_curr.append(util.getRelPath(out_file_curr,'/disk3'));
# # print dir_curr.split('/');
# captions_curr.append(dir_curr.split('/')[-2]+' '+dir_curr.split('/')[-1]+' '+key_curr);
# img_paths.append(img_paths_curr);
# captions.append(captions_curr);
# visualize.writeHTML(out_file_html,img_paths,captions,height=200,width=200);
# print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
return
dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug','/disk3/maheen_data/headC_160/noFlow_human_debug'];
out_file_html=os.path.join(dirs[0],'comparison_dloss_seg_score.html');
dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct']];
out_dirs=[os.path.join(dir_curr,'plots') for dir_curr in dirs];
file_seg_pre='dloss_seg_';
file_score_pre='dloss_score_';
num_iters=range(2,21);
model_num=range(25000,100000,20000);
model_num.append(100000);
model_num=[str(model_num_curr) for model_num_curr in model_num]
img_paths=[];
captions=[];
for dir_curr,out_dir_curr in zip(dirs,out_dirs):
dict_for_plotting={'seg_all':OrderedDict(),'score_all':OrderedDict()};
for model_num_curr in model_num:
seg_all=[];
score_all=[];
for num_iter_curr in num_iters:
file_curr_seg=os.path.join(dir_curr,model_num_curr,file_seg_pre+str(num_iter_curr)+'.npy');
file_curr_score=os.path.join(dir_curr,model_num_curr,file_score_pre+str(num_iter_curr)+'.npy');
seg_curr=np.load(file_curr_seg);
seg_curr=np.unique(np.ravel(seg_curr));
score_curr=list(np.load(file_curr_score));
score_curr=list(np.unique(np.ravel(score_curr)));
seg_all.extend(seg_curr);
score_all.extend(score_curr);
seg_all=list(set(seg_all));
score_all=list(set(score_all));
seg_all.sort();
score_all.sort();
dict_for_plotting['seg_all'][model_num_curr]=seg_all;
dict_for_plotting['score_all'][model_num_curr]=score_all;
img_paths_curr=[];
captions_curr=[];
for key_curr in dict_for_plotting.keys():
out_file_curr=os.path.join(out_dir_curr,key_curr+'.png');
data=dict_for_plotting[key_curr];
xAndYs=data.values();
legend_entries=data.keys();
xAndYs=[(range(len(x_curr)),x_curr) for x_curr in xAndYs];
visualize.plotSimple(xAndYs,out_file_curr,title=key_curr,xlabel='sorted idx',ylabel='values',legend_entries=legend_entries,outside=True);
print out_file_curr.replace('/disk3','vision3.cs.ucdavis.edu:1001');
img_paths_curr.append(util.getRelPath(out_file_curr,'/disk3'));
# print dir_curr.split('/');
captions_curr.append(dir_curr.split('/')[-2]+' '+dir_curr.split('/')[-1]+' '+key_curr);
img_paths.append(img_paths_curr);
captions.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths,captions,height=200,width=200);
print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
# dloss_seg=np.load(file_curr_seg);
# dloss_seg=np.mean(dloss_seg,axis=0);
# print dloss_seg.shape;
# dloss_seg=dloss_seg[0];
# print dloss_seg.shape;
# dloss_score=np.load(file_curr_score);
# print dloss_score.shape;
# print np.min(dloss_score);
# print np.max(dloss_score);
# print np.min(dloss_seg);
# print np.max(dloss_seg);
# # print dloss_seg[0],np.min(dloss_seg),np.max(dloss_seg);
# print file_curr;
return
dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug','/disk3/maheen_data/headC_160/noFlow_human_debug'];
out_file_html=os.path.join('/disk3/maheen_data/headC_160/withFlow_human_debug','comparison_grads_weights_ratios.html');
# dirs=['/disk3/maheen_data/headC_160/withFlow_human_debug'];
out_file_html=os.path.join('/disk3/maheen_data/headC_160/withFlow_human_debug','comparison_grads_seg_no_seg.html');
dirs=[os.path.join(dir_curr,dir_in) for dir_curr in dirs for dir_in in ['correct','incorrect']];
out_dirs=[os.path.join(dir_curr,'plots') for dir_curr in dirs];
[util.mkdir(out_dir_curr) for out_dir_curr in out_dirs];
model_num=range(5000,100000,20000);
model_num.append(100000);
model_num=[str(model_num_curr) for model_num_curr in model_num]
num_iters=range(1,21);
file_pre_weight='weight_mag_';
file_pre_grad='grad_mag_';
# file_curr=os.path.join(dirs[0],model_num[-1],file_pre_grad+'1.npy');
# grads=np.load(file_curr);
# print grads.shape;
# grads=grads[::2];
# print grads.shape;
# print grads[26:]
# compareMagInfoLayerTime(dirs,out_dirs,model_num,num_iters,file_pre_grad,file_pre_weight,out_file_html)
# layer_range=range(26,32);
layer_range=[26,27,28,31];
num_iters=range(1,21,2);
img_paths_seg_flow=compareGradInfoLayer(dirs[:2],out_dirs[:2],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
layer_range=range(26,32);
num_iters=range(2,21,2);
img_paths_score_flow=compareGradInfoLayer(dirs[:2],out_dirs[:2],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
# layer_range=range(13,18)
layer_range=[13,14,17]
num_iters=range(1,21,2);
img_paths_seg_noflow=compareGradInfoLayer(dirs[2:],out_dirs[2:],model_num,num_iters,file_pre_grad,layer_range,'seg_grad_mag')
layer_range=range(13,18);
num_iters=range(2,21,2);
img_paths_score_noflow=compareGradInfoLayer(dirs[2:],out_dirs[2:],model_num,num_iters,file_pre_grad,layer_range,'score_grad_mag')
img_paths=[img_paths_seg_flow,img_paths_score_flow,img_paths_seg_noflow,img_paths_score_noflow];
img_paths=[[util.getRelPath(path_curr,'/disk3') for path_curr in list_curr] for list_curr in img_paths];
# print img_paths
captions=[];
# path='../../../../../../..//maheen_data/headC_160/noFlow_human_debug/incorrect/plots/score_grad_mag.png'
for list_curr in img_paths:
captions_curr=[];
for path in list_curr:
path_split=path.split('/');
caption=path_split[-4]+' '+path_split[-3];
captions_curr.append(caption);
print caption
captions.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths,captions,height=300,width=300);
print out_file_html.replace('/disk3','vision3.cs.ucdavis.edu:1001');
# out_files=compareGradInfoLayer(dirs,out_dirs,model_num,num_iters,file_pre_grad,layer_range)
# break;
return
params=[('/disk3/maheen_data/headC_160/withFlow_gaussian_human_softmax/intermediate','/disk3/maheen_data/headC_160/withFlow_human_debug/correct','40'),
('/disk3/maheen_data/headC_160/withFlow_gaussian_human/intermediate','/disk3/maheen_data/headC_160/withFlow_human_debug/incorrect','56')];
out_file_commands_pre='/disk3/maheen_data/headC_160/withFlow_human_debug/debug_commands_';
path_to_train_file='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_withFlow_debug.th'
# params=[('/disk3/maheen_data/headC_160/noFlow_gaussian_human_softmax/intermediate_res','/disk3/maheen_data/headC_160/noFlow_human_debug/correct','40'),
# ('/disk3/maheen_data/headC_160/noFlow_gaussian_human/intermediate','/disk3/maheen_data/headC_160/noFlow_human_debug/incorrect','56')];
# [util.mkdir(params_curr[1]) for params_curr in params];
# out_file_commands_pre='/disk3/maheen_data/headC_160/noFlow_human_debug/debug_commands_';
# path_to_train_file='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_noFlow_debug.th'
model_num=range(5000,100000,20000);
model_num.append(100000);
writeCommandsForTrainDebug(params,path_to_train_file,out_file_commands_pre,model_num)
return
print 'hello';
dir_debug_with_flow='/disk3/maheen_data/headC_160/withFlow_human_debug';
dir_debug_no_flow='/disk3/maheen_data/headC_160/noFlow_human_debug';
score_dir='score_gradient_start';
seg_dir='seg_gradient_start';
dirs=[os.path.join(dir_debug_no_flow,score_dir),os.path.join(dir_debug_with_flow,score_dir),
os.path.join(dir_debug_no_flow,seg_dir),os.path.join(dir_debug_with_flow,seg_dir)]
for dir_curr in dirs:
np_files=util.getFilesInFolder(dir_curr,'.npy');
np_nums=[int(file_curr[file_curr.rindex('_')+1:file_curr.rindex('.')]) for file_curr in np_files];
sort_idx=np.argsort(np_nums);
np_files=np.array(np_files)[sort_idx];
gradients=getGradientMags(np_files);
print dir_curr;
print len(gradients),np.mean(gradients),min(gradients),max(gradients);
def main():
out_dir='/disk2/mayExperiments/eval_nC_zS_youtube';
model_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/opt_noFix_conv1_conv2_conv3_conv4_conv5_llr__iter_50000.caffemodel';
clusters_file='/disk3/maheen_data/youtube_train_40/clusters_100000.mat';
flo_dir_meta=os.path.join(out_dir,'ft_youtube_model')
flo_dir=os.path.join(flo_dir_meta,'flo');
match_info_file=os.path.join(flo_dir,'match_info.txt');
train_val_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/train_val_conv1_conv2_conv3_conv4_conv5.prototxt';
out_dir_flo_viz_org=os.path.join(flo_dir_meta,'flo_viz');
gpu=0;
h5_files,img_files,img_sizes=po.parseInfoFile(match_info_file);
file_names=util.getFileNames(img_files,ext=False)
out_dirs_flo_viz=[out_dir_flo_viz_org];
n_range=[5,10];
n_range_str='_'.join([str(n) for n in n_range]);
out_file_html=os.path.join(out_dir,'flo_n_'+n_range_str+'.html');
for n in n_range:
out_dir_flo=os.path.join(flo_dir_meta,'flo_n_'+str(n));
out_dir_flo_viz=os.path.join(flo_dir_meta,'flo_n_'+str(n)+'_viz');
out_file_sh=out_dir_flo_viz+'.sh'
util.mkdir(out_dir_flo);
util.mkdir(out_dir_flo_viz);
out_dirs_flo_viz.append(out_dir_flo_viz)
flo_files=[os.path.join(out_dir_flo,file_name+'.flo') for file_name in file_names];
for h5_file,img_size,flo_file in zip(h5_files,img_sizes,flo_files):
flow_resize=getMatFromH5TopN(h5_file,img_size,clusters_file,n);
util.writeFlowFile(flow_resize,flo_file);
out_files_viz=[os.path.join(out_dir_flo_viz,file_name+'.png') for file_name in file_names];
po.writeScriptToGetFloViz(flo_files,out_files_viz,out_file_sh);
subprocess.call('sh '+out_file_sh,shell=True);
img_paths_html=[];
captions_html=[];
for img_file,file_name in zip(img_files,file_names):
row_curr=[img_file];
caption_curr=[file_name];
for out_dir_flo_curr in out_dirs_flo_viz:
file_curr=os.path.join(out_dir_flo_curr,file_name+'.png');
row_curr.append(file_curr);
caption_curr.append(util.getFileNames([out_dir_flo_curr])[0]);
row_curr=[util.getRelPath(f) for f in row_curr];
img_paths_html.append(row_curr);
captions_html.append(caption_curr);
util.writeHTML(out_file_html,img_paths_html,captions_html);
print out_file_html.replace('/disk2','vision3.cs.ucdavis.edu:1000/');
print 'hello';
def main():
out_dir_training_files='/disk3/maheen_data/flo_only_training_files';
# model='/disk2/aprilExperiments/headC_160/withFlow_gaussian_xavier_fixed_unitCombo_floStumpPretrained.dat'
# out_dir='/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_floStumpPretrained_fullTraining'
model='/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_floStumpPretrained_fullTraining/intermediate/model_all_45000.dat'
out_dir='/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_floStumpPretrained_fullTraining_res'
util.mkdir(out_dir);
out_file_pos_train=os.path.join(out_dir_training_files,'pos_human_flo_train.txt');
out_file_neg_train=os.path.join(out_dir_training_files,'neg_human_flo_train.txt');
out_file_pos_val=os.path.join(out_dir_training_files,'pos_human_flo_val.txt');
out_file_neg_val=os.path.join(out_dir_training_files,'neg_human_flo_val.txt');
out_file_neg='/disk3/maheen_data/headC_160/neg_flos/negatives_onlyHuman_withFlow.txt'
out_file_pos='/disk3/maheen_data/headC_160/noFlow_gaussian_human/pos_flos/positives_onlyHuman_withFlow.txt'
# shortenTrainingData(out_file_pos,out_file_pos_train,0.9,val_txt_new=out_file_pos_val)
# shortenTrainingData(out_file_neg,out_file_neg_train,0.9,val_txt_new=out_file_neg_val)
learningRate=0.00001;
path_to_th='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_withFlow_cl.th';
iterations=55000;
out_log_curr=os.path.join(out_dir,'log.txt');
out_im_pre=os.path.join(out_dir,'loss');
command_curr='python script_visualizeLoss.py -log_file '+'/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_floStumpPretrained_fullTraining/log.txt'+' '+out_log_curr+' -out_file_pre '+out_im_pre+' -val ';
print command_curr;
os.system(command_curr);
return
command=printCommandToTrain(path_to_th,model,out_dir,iterations=iterations,learningRate=learningRate,pos_file=out_file_pos_train,
pos_val_file=out_file_pos_val,neg_file=out_file_neg_train,neg_val_file=out_file_neg_val,testAfter=40);
# util.writeFile(os.path.join(out_dir,'train_command.sh'),[command]);
return
out_dir_training_files='/disk3/maheen_data/flo_only_training_files';
num=64;
out_file_neg='/disk3/maheen_data/headC_160/neg_flos/negatives_onlyHuman_withFlow.txt'
out_file_pos='/disk3/maheen_data/headC_160/noFlow_gaussian_human/pos_flos/positives_onlyHuman_withFlow.txt'
out_file_pos_mini=os.path.join(out_dir_training_files,'pos_human_withflo_'+str(num)+'.txt')
out_file_neg_mini=os.path.join(out_dir_training_files,'neg_human_withflo_'+str(num)+'.txt');
# shortenTrainingData(out_file_pos,out_file_pos_mini,num);
# shortenTrainingData(out_file_neg,out_file_neg_mini,num);
path_to_th='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_withFlow_cl.th';
model='/disk2/aprilExperiments/headC_160/withFlow_gaussian_xavier_fixed.dat';
out_dir='/disk3/maheen_data/headC_160/withFlow_human_xavier_unit_lr_search';
out_file_html=os.path.join(out_dir,'comparison.html');
out_file_sh = os.path.join(out_dir,'lr_search.sh');
util.mkdir(out_dir);
iterations=1000;
commands=[];
img_paths=[];
captions=[];
for learningRate in [0.001,0.0001,0.00001,0.000001, 0.0000001]:
out_dir_curr=os.path.join(out_dir,str(learningRate));
util.mkdir(out_dir_curr);
out_log_curr=os.path.join(out_dir_curr,'log.txt');
out_im_pre=os.path.join(out_dir_curr,'loss');
command_curr='python script_visualizeLoss.py -log_file '+out_log_curr+' -out_file_pre '+out_im_pre;
print command_curr;
os.system(command_curr);
img_paths_curr=[util.getRelPath(out_im_pre+post,'/disk3') for post in ['_score.png','_seg.png']];
captions_curr=[str(learningRate)+' '+file_curr for file_curr in util.getFileNames(img_paths_curr)];
captions.append(captions_curr);
img_paths.append(img_paths_curr);
# command=printCommandToTrain(path_to_th,model,out_dir_curr,iterations=iterations,learningRate=learningRate,pos_file=out_file_pos_mini,
# pos_val_file=out_file_pos_mini,neg_file=out_file_neg_mini,neg_val_file=out_file_neg_mini);
# commands.append(command);
# break;
# util.writeFile(out_file_sh,commands);
# print 'sh '+out_file_sh
visualize.writeHTML(out_file_html,img_paths,captions);
return
out_dir_training_files='/disk3/maheen_data/flo_only_training_files';
out_dir='/disk3/maheen_data/headC_160/onlyFlow_human_xavier_fix_full';
out_file_html=os.path.join(out_dir,'comparison_loss.html');
out_file_sh=os.path.join(out_dir,'lr_commands.sh');
util.mkdir(out_dir_training_files);
out_file_pos=os.path.join(out_dir_training_files,'pos_human_only_flo.txt');
out_file_neg=os.path.join(out_dir_training_files,'neg_human_only_flo.txt');
out_file_pos_train=os.path.join(out_dir_training_files,'pos_human_only_flo_train.txt');
out_file_neg_train=os.path.join(out_dir_training_files,'neg_human_only_flo_train.txt');
out_file_pos_val=os.path.join(out_dir_training_files,'pos_human_only_flo_val.txt');
out_file_neg_val=os.path.join(out_dir_training_files,'neg_human_only_flo_val.txt');
# shortenTrainingData(out_file_pos,out_file_pos_train,0.9,val_txt_new=out_file_pos_val)
# shortenTrainingData(out_file_neg,out_file_neg_train,0.9,val_txt_new=out_file_neg_val)
path_to_th='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_noFlow_cl.th';
model='/disk3/maheen_data/headC_160/onlyFlow_human_xavier_fix_full/1e-05/intermediate/model_all_35000.dat';
# out_file_html=os.path.join(out_dir,'loss_comparison.html');
iterations=65000;
testAfter=40;
util.mkdir(out_dir);
commands=[];
img_paths=[];
captions=[];
for learningRate in [0.000001]:
out_dir_curr=os.path.join(out_dir,str(learningRate*10)+'_res');
util.mkdir(out_dir_curr);
# command_curr='python script_visualizeLoss.py '+os.path.join(out_dir_curr,'log.txt')+' '+os.path.join(out_dir_curr,'loss');
# os.system(command_curr);
# img_paths.append([util.getRelPath(os.path.join(out_dir_curr,'loss'+post_curr),'/disk3') for post_curr in ['_seg.png','_score.png']]);
# captions.append([str(learningRate)+' '+file_name for file_name in util.getFileNames(img_paths[-1])]);
command=printCommandToTrain(path_to_th,model,out_dir_curr,iterations=iterations,learningRate=learningRate,pos_file=out_file_pos_train,
pos_val_file=out_file_pos_val,neg_file=out_file_neg_train,neg_val_file=out_file_neg_val,testAfter=testAfter);
print command
# commands.append(command);
# util.writeFile(out_file_sh,commands);
# print out_file_sh
return
out_dir_training_files='/disk3/maheen_data/flo_only_training_files';
out_dir='/disk3/maheen_data/headC_160/onlyFlow_human_xavier_fix_full';
out_file_html=os.path.join(out_dir,'comparison_loss.html');
out_file_sh=os.path.join(out_dir,'lr_commands.sh');
util.mkdir(out_dir_training_files);
out_file_pos=os.path.join(out_dir_training_files,'pos_human_only_flo.txt');
out_file_neg=os.path.join(out_dir_training_files,'neg_human_only_flo.txt');
out_file_pos_train=os.path.join(out_dir_training_files,'pos_human_only_flo_train.txt');
out_file_neg_train=os.path.join(out_dir_training_files,'neg_human_only_flo_train.txt');
out_file_pos_val=os.path.join(out_dir_training_files,'pos_human_only_flo_val.txt');
out_file_neg_val=os.path.join(out_dir_training_files,'neg_human_only_flo_val.txt');
# shortenTrainingData(out_file_pos,out_file_pos_train,0.9,val_txt_new=out_file_pos_val)
# shortenTrainingData(out_file_neg,out_file_neg_train,0.9,val_txt_new=out_file_neg_val)
path_to_th='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_noFlow_cl.th';
model='/disk2/aprilExperiments/headC_160/noFlow_gaussian_xavier_fixed.dat';
out_file_html=os.path.join(out_dir,'loss_comparison.html');
iterations=100000;
testAfter=40;
util.mkdir(out_dir);
commands=[];
img_paths=[];
captions=[];
for learningRate in [0.00001,0.000001]:
out_dir_curr=os.path.join(out_dir,str(learningRate));
util.mkdir(out_dir_curr);
command_curr='python script_visualizeLoss.py '+os.path.join(out_dir_curr,'log.txt')+' '+os.path.join(out_dir_curr,'loss');
os.system(command_curr);
img_paths.append([util.getRelPath(os.path.join(out_dir_curr,'loss'+post_curr),'/disk3') for post_curr in ['_seg.png','_score.png']]);
captions.append([str(learningRate)+' '+file_name for file_name in util.getFileNames(img_paths[-1])]);
# command=printCommandToTrain(path_to_th,model,out_dir_curr,iterations=iterations,learningRate=learningRate,pos_file=out_file_pos_train,
# pos_val_file=out_file_pos_val,neg_file=out_file_neg_train,neg_val_file=out_file_neg_val,testAfter=testAfter);
# commands.append(command);
# util.writeFile(out_file_sh,commands);
# print out_file_sh
visualize.writeHTML(out_file_html,img_paths,captions,500,500);
return
num=64;
out_file_pos_mini=os.path.join(out_dir_training_files,'pos_human_only_flo_'+str(num)+'.txt')
out_file_neg_mini=os.path.join(out_dir_training_files,'neg_human_only_flo_'+str(num)+'.txt');
# shortenTrainingData(out_file_pos,out_file_pos_mini,num);
# shortenTrainingData(out_file_neg,out_file_neg_mini,num);
path_to_th='/home/maheenrashid/Downloads/deep_proposals/torch_new/headC_160_noFlow_cl.th';
model='/disk2/aprilExperiments/headC_160/noFlow_gaussian_xavier_fixed_unitCombo.dat';
iterations=1000;
util.mkdir(out_dir);
commands=[];
img_paths=[];
captions=[];
for learningRate in [0.001,0.0001,0.00001,0.000001, 0.0000001]:
out_dir_curr=os.path.join(out_dir,str(learningRate));
util.mkdir(out_dir_curr);
out_log_curr=os.path.join(out_dir_curr,'log.txt');
out_im_pre=os.path.join(out_dir_curr,'loss');
# command_curr='python script_visualizeLoss.py '+out_log_curr+' '+out_im_pre;
# print command_curr;
# os.system(command_curr);
# img_paths_curr=[util.getRelPath(out_im_pre+post,'/disk3') for post in ['_score.png','_seg.png']];
# captions_curr=[str(learningRate)+' '+file_curr for file_curr in util.getFileNames(img_paths_curr)];
# captions.append(captions_curr);
# img_paths.append(img_paths_curr);
command=printCommandToTrain(path_to_th,model,out_dir_curr,iterations=iterations,learningRate=learningRate,pos_file=out_file_pos_mini,
pos_val_file=out_file_pos_mini,neg_file=out_file_neg_mini,neg_val_file=out_file_neg_mini);
print command;
commands.append(command);
# util.writeFile(out_file_sh,commands);
# print out_file_sh
# visualize.writeHTML(out_file_html,img_paths,captions);
return
dir_curr='/disk3/maheen_data/headC_160/withFlow_xavier_16_score/intermediate';
range_files=range(2000,96000,2000);
to_del=range(4000,94000,4000);
for model_num in to_del:
file_curr=os.path.join(dir_curr,'model_all_'+str(model_num)+'.dat');
os.remove(file_curr);
print file_curr;
# assert os.path.exists(file_curr);
return
out_dir='/disk3/maheen_data/flo_only_training_files';
util.mkdir(out_dir);
out_file_pos=os.path.join(out_dir,'pos_human_only_flo.txt');
out_file_neg=os.path.join(out_dir,'neg_human_only_flo.txt');
neg_flo='/disk3/maheen_data/headC_160/neg_flos/negatives_onlyHuman_withFlow.txt';
pos_flo='/disk3/maheen_data/headC_160/noFlow_gaussian_human/pos_flos/positives_onlyHuman_withFlow.txt';
convertFileToFloOnly(neg_flo,out_file_neg)
convertFileToFloOnly(pos_flo,out_file_pos);
return
pos_all=util.readLinesFromFile(pos_all);
pos_flo=util.readLinesFromFile(pos_flo);
neg_all=util.readLinesFromFile(neg_all);
neg_flo=util.readLinesFromFile(neg_flo);
print pos_all[0];
print pos_flo[0];
print '___';
print neg_all[0];
print neg_flo[0];
return
pos_human='/disk2/aprilExperiments/positives_160.txt';
neg_human='/disk2/marchExperiments/deep_proposals/negatives.txt';
pos_human_small='/disk2/aprilExperiments/positives_160_oneHundreth.txt';
neg_human_small='/disk2/marchExperiments/deep_proposals/negatives_oneHundreth.txt';
ratio_txt=100;
shortenTrainingData(pos_human,pos_human_small,ratio_txt);
shortenTrainingData(neg_human,neg_human_small,ratio_txt);
return
files=['/disk3/maheen_data/headC_160/withFlow_human_debug/correct/100000/im_layer_weights_1.npy',
'/disk3/maheen_data/headC_160/withFlow_human_debug/correct/100000/flo_layer_weights_1.npy',
'/disk3/maheen_data/headC_160/withFlow_human_debug/incorrect/100000/im_layer_weights_1.npy',
'/disk3/maheen_data/headC_160/withFlow_human_debug/incorrect/100000/flo_layer_weights_1.npy'];
for file_curr in files:
out_dir_curr=os.path.join(file_curr.rsplit('/',2)[0],'plots');
weights=np.load(file_curr);
# print weights.shape;
# weights=weights[:,:,:2,:];
# print weights.shape
# weights=weights.transpose((0,2,3,1));
# print weights.shape
data=vis_square(weights);
# print data.shape
out_file=os.path.join(out_dir_curr,util.getFileNames([file_curr],ext=False)[0]+'.png')
print out_file;
scipy.misc.imsave(out_file,data);
def main():
# data='/home/SSD3/maheen-data/horse_project/data_resize/horse/matches_5_train_allKP.txt'
# # /home/SSD3/maheen-data/horse_project/data_resize/horse/matches_5_train_allKP.txt
# to_search=\
# ['/home/SSD3/maheen-data/horse_project/data_check/horse/im/horses_pascal_selected/2009_004662.jpg /home/SSD3/maheen-data/horse_project/data_check/horse/npy/horses_pascal_selected/2009_004662.npy',
# '/home/SSD3/maheen-data/horse_project/data_check/horse/im/imagenet_n02374451/n02374451_11539.jpg /home/SSD3/maheen-data/horse_project/data_check/horse/npy/imagenet_n02374451/n02374451_11539.npy',
# '/home/SSD3/maheen-data/horse_project/data_check/horse/im/imagenet_n02374451/n02374451_16786.jpg /home/SSD3/maheen-data/horse_project/data_check/horse/npy/imagenet_n02374451/n02374451_16786.npy',
# '/home/SSD3/maheen-data/horse_project/data_check/horse/im/imagenet_n02374451/n02374451_4338.jpg /home/SSD3/maheen-data/horse_project/data_check/horse/npy/imagenet_n02374451/n02374451_4338.npy']
# data=util.readLinesFromFile(data);
# print data[0];
# to_search=[file_curr.replace('data_check','data_resize') for file_curr in to_search];
# idx_lines=[data.index(line_curr) for line_curr in to_search if line_curr in data];
# print idx_lines;
# for idx_line_curr in idx_lines:
# print 'batch_no',(idx_line_curr)/64
# # npy_files=[file_curr[file_curr.index(' ')+1:] for file_curr in data];
# # print npy_files[0];
# # print len(npy_files);
# # p=multiprocessing.Pool(multiprocessing.cpu_count());
# # problem_files=p.map(findProblemNPYMP,npy_files);
# # problem_files=[file_curr for file_curr in problem_files if file_curr is not None];
# # print (len(problem_files));
# return
# data='/home/laoreja/new-deep-landmark/train/vanilla/aflw_224/aflw_vanilla_val_224.txt';
# data='/home/laoreja/new-deep-landmark/train/vanilla/aflw_224/aflw_vanilla_train_224_weight.txt';
# data=util.readLinesFromFile(data);
# print data;
# total=0;
# for h5_file_curr in data:
# with h5py.File(h5_file_curr,'r') as hf:
# print('List of arrays in this file: ', hf.keys())
# data = hf.get('confidence')
# np_data = np.array(data)
# total=total+np_data.shape[0];
# print('Shape of the array dataset_1: ', np_data.shape)
# print total;
# return
# horse_path='/home/SSD3/maheen-data/horse_project/data_resize/horse/matches_5_train_allKP.txt'
# human_path_noIm='/home/SSD3/maheen-data/horse_project/data_resize/aflw/matches_5_train_allKP_noIm.txt'
# human_path='/home/SSD3/maheen-data/horse_project/data_resize/aflw/matches_5_train_allKP.txt'
# paths=[horse_path,human_path_noIm,human_path];
# out_files=[file_curr[:file_curr.rindex('.')]+'_dummy.txt' for file_curr in paths];
# for file_curr,out_file_curr in zip(paths,out_files):
# data_curr=util.readLinesFromFile(file_curr);
# data_curr=data_curr[0:50:5];
# # print data_curr;
# print len(data_curr);
# util.writeFile(out_file_curr,data_curr);
# print out_file_curr;
# return
# im_path= "/home/SSD3/maheen-data/horse_project/data_resize/horse/im/_04_Aug16_png/horse+head12.jpg"
# # 2 : "/home/SSD3/maheen-data/horse_project/data_resize/horse/npy/_04_Aug16_png/horse+head12.npy"
# # "/home/SSD3/maheen-data/horse_project/data_resize/aflw/im/0/image67102_20650.jpg"
# np_path="/home/SSD3/maheen-data/horse_project/data_resize/horse/npy/_04_Aug16_png/horse+head12.npy"
# # "/home/SSD3/maheen-data/horse_project/data_resize/aflw/npy/0/image67102_20650.npy"
# # im=scipy.misc.read(im_path);
# im=cv2.imread(im_path);
# labels=np.load(np_path);
# print labels
# for i in xrange(labels.shape[0]):
# cv2.circle(im, (labels[i][0], labels[i][1]), 2, (0,0,255), -1)
# cv2.imwrite('/home/SSD3/maheen-data/temp/check.png', im)
# return
# path_to_th='/home/maheenrashid/Downloads/horses/torch/test_tps_cl.th';
# iterations=10;
# out_dir_models='/home/SSD3/maheen-data/horse_human_fiveKP_tps_adam'
# model_pre=os.path.join(out_dir_models,'intermediate','model_all_');
# model_post='.dat';
# range_models=range(450,4500+1,450);
# out_dir_meta=os.path.join(out_dir_models,'test_overtime');
# batch_size=60;
# # commands=generateTPSTestCommands(path_to_th,batch_size,iterations,model_pre,model_post,range_models,out_dir_meta)
# # print len(commands);
# # print commands[0];
# # out_file_commands=os.path.join(out_dir_meta+'.sh');
# # util.writeFile(out_file_commands,commands);
# dir_server='/home/SSD3/maheen-data';
# range_batches=range(1,10);
# # batch_size=60;
# range_images=range(1,61,5);
# img_dir_meta='/home/SSD3/maheen-data/horse_human_fiveKP_tps_adam/test_overtime'
# img_dir=[os.path.join(img_dir_meta,'model_all_'+str(range_model_curr)) for range_model_curr in range_models]
# out_file_html='/home/SSD3/maheen-data/horse_human_fiveKP_tps_adam/test_viz.html'
# file_post=['_horse.jpg','_human.jpg','_gtwarp.jpg','_predwarp.jpg']
# loss_post='_loss.npy';
# out_file_html=img_dir_meta+'.html';
# img_caption_pre=[str(model_num) for model_num in range_models];
# comparativeLossViz(img_dir,file_post,loss_post,range_batches,range_images,out_file_html,dir_server,img_caption_pre)
# return
dir_server = '/home/SSD3/maheen-data'
range_batches = range(1, 9)
# batch_size=60;
range_images = range(1, 129, 5)
img_dir = ['/home/SSD3/maheen-data/horse_human_fiveKP_tps_adam/test_viz/']
# out_file_html='/home/SSD3/maheen-data/horse_human_fiveKP_tps_adam/test_viz.html'
img_dir = [
'/home/SSD3/maheen-data/horse_project/tps_train_allKP_adam/test_viz'
]
out_file_html = '/home/SSD3/maheen-data/horse_project/tps_train_allKP_adam/test_viz.html'
file_post = ['_horse.jpg', '_human.jpg', '_gtwarp.jpg', '_predwarp.jpg']
loss_post = '_loss.npy'
comparativeLossViz(img_dir, file_post, loss_post, range_batches,
range_images, out_file_html, dir_server)
return
img_files = []
caption_files = []
out_dir = '/home/SSD3/maheen-data/training_kp_withWarp_test_debug_tps_adam'
out_dir = '/home/SSD3/maheen-data/testing_5_kp_withWarp_fixed_adam_debug'
out_dir = '/home/SSD3/maheen-data/training_5_kp_withWarp_fixed_adam__1e-05/test'
dir_server = '/home/SSD3/maheen-data'
out_file_html = os.path.join(out_dir, 'viz.html')
for i in range(1, 94):
im_file = os.path.join(out_dir,
str(i) + '_org.jpg')
warp_file = os.path.join(out_dir,
str(i) + '_warp.jpg')
im_file_small = os.path.join(out_dir,
str(i) + '_small_org.jpg')
warp_file_small = os.path.join(out_dir,
str(i) + '_small_warp.jpg')
im_file = util.getRelPath(im_file, dir_server)
warp_file = util.getRelPath(warp_file, dir_server)
im_file_small = util.getRelPath(im_file_small, dir_server)
warp_file_small = util.getRelPath(warp_file_small, dir_server)
img_files.append([im_file, warp_file])
# ,im_file_small,warp_file_small]);
caption_files.append([str(i) + ' org',
str(i) + ' warp'])
# ,'small_org','small_warp']);
visualize.writeHTML(out_file_html, img_files, caption_files, 224, 224)
return
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
num_neighbors = 5
out_file_human = os.path.join(
out_dir_meta_face, 'matches_' + str(num_neighbors) + '_val_fiveKP.txt')
out_file_human_new = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_val_fiveKP_noIm.txt')
modifyHumanFile(out_file_human, out_file_human_new)
# out_dir_meta_face='/home/SSD3/maheen-data/horse_project/aflw';
out_file_human = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_train_fiveKP.txt')
out_file_human_new = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_train_fiveKP_noIm.txt')
modifyHumanFile(out_file_human, out_file_human_new)
return
# matches_file='/home/maheenrashid/Downloads/knn_5_points_train_list_clean.txt'
# face_data_file='/home/laoreja/new-deep-landmark/dataset/train/aflw_trainImageList.txt';
# # face_data_file_old='/home/laoreja/deep-landmark-master/dataset/train/trainImageList.txt';
# face_data_list_file='/home/SSD3/maheen-data/aflw_data/npy/data_list.txt';
# out_dir_meta_horse='/home/SSD3/maheen-data/horse_project/horse';
# out_dir_meta_horse_list=[os.path.join(out_dir_meta_horse,'im'),os.path.join(out_dir_meta_horse,'npy')];
# out_dir_meta_face='/home/SSD3/maheen-data/horse_project/aflw';
# out_dir_meta_face_list=[os.path.join(out_dir_meta_face,'im'),os.path.join(out_dir_meta_face,'npy')];
# out_dir_meta_face_old='/home/SSD3/maheen-data/horse_project/face';
# out_dir_meta_face_old_list=[os.path.join(out_dir_meta_face_old,'im'),os.path.join(out_dir_meta_face_old,'npy')];
# num_neighbors=5;
# out_file_face=os.path.join(out_dir_meta_face,'matches_'+str(num_neighbors)+'_train_fiveKP.txt');
# out_file_horse=os.path.join(out_dir_meta_horse,'matches_'+str(num_neighbors)+'_train_fiveKP.txt');
# missing_files=makeMatchFile(num_neighbors,matches_file,face_data_file,out_dir_meta_horse_list,out_dir_meta_face_list,out_file_horse,out_file_face,out_dir_meta_face_old_list)
# return
# out_dir_meta_face='/home/SSD3/maheen-data/horse_project/aflw';
# num_neighbors=5;
# out_file_human=os.path.join(out_dir_meta_face,'matches_'+str(num_neighbors)+'_val_fiveKP.txt');
# out_file_human_new=os.path.join(out_dir_meta_face,'matches_'+str(num_neighbors)+'_val_fiveKP_noIm.txt');
# # modifyHumanFile(out_file_human,out_file_human_new)
# # out_dir_meta_face='/home/SSD3/maheen-data/horse_project/aflw';
# out_file_human=os.path.join(out_dir_meta_face,'matches_'+str(num_neighbors)+'_train_fiveKP.txt');
# out_file_human_new=os.path.join(out_dir_meta_face,'matches_'+str(num_neighbors)+'_train_fiveKP_noIm.txt');
# # modifyHumanFile(out_file_human,out_file_human_new)
# print out_file_human_new;
# return
# img_files=[];
# caption_files=[];
# out_dir='/home/SSD3/maheen-data/training_kp_withWarp_test_final'
# dir_server='/home/SSD3/maheen-data';
# out_file_html=os.path.join(out_dir,'viz.html');
# for i in range(1,94):
# im_file=os.path.join(out_dir,str(i)+'.jpg');
# warp_file=os.path.join(out_dir,str(i)+'_warp.jpg');
# im_file=util.getRelPath(im_file,dir_server);
# warp_file=util.getRelPath(warp_file,dir_server);
# img_files.append([im_file,warp_file]);
# caption_files.append(['org','warp']);
# visualize.writeHTML(out_file_html,img_files,caption_files,224,224);
# return
file_horse = '/home/SSD3/maheen-data/horse_project/horse/matches_5_train_fiveKP.txt'
out_file_horse = '/home/SSD3/maheen-data/horse_project/horse_resize/matches_5_train_fiveKP.txt'
lines = util.readLinesFromFile(file_horse)
print len(lines)
lines = list(set(lines))
print len(lines)
lines = [line_curr.split(' ') for line_curr in lines]
im_files = [line_curr[0] for line_curr in lines]
npy_files = [line_curr[1] for line_curr in lines]
out_dir_meta_old = '/home/SSD3/maheen-data/horse_project/horse/'
out_dir_meta_new = '/home/SSD3/maheen-data/horse_project/horse_resize/'
replace_paths = [out_dir_meta_old, out_dir_meta_new]
args = []
for idx in range(len(im_files)):
im_file = im_files[idx]
npy_file = npy_files[idx]
out_im_file = im_file.replace(replace_paths[0], replace_paths[1])
out_npy_file = npy_file.replace(replace_paths[0], replace_paths[1])
args.append((idx, im_file, npy_file, out_im_file, out_npy_file))
p = multiprocessing.Pool(multiprocessing.cpu_count())
p.map(resizeImAndNpy224, args)
out_dir_meta_old = '/home/SSD3/maheen-data/horse_project/horse/'
out_dir_meta_new = '/home/SSD3/maheen-data/horse_project/horse_resize/'
replace_paths = [out_dir_meta_old, out_dir_meta_new]
lines = util.readLinesFromFile(file_horse)
lines_new = [
line.replace(replace_paths[0], replace_paths[1]) for line in lines
]
util.writeFile(out_file_horse, lines_new)
lines = util.readLinesFromFile(out_file_horse)
print(len(lines))
im_file = lines[90].split(' ')[0]
im = cv2.imread(im_file, 1)
labels = np.load(lines[90].split(' ')[1])
for i in xrange(labels.shape[0]):
cv2.circle(im, (labels[i][0], labels[i][1]), 2, (0, 0, 255), -1)
cv2.imwrite('/home/SSD3/maheen-data/temp/check.png', im)
return
dir_out = '/home/SSD3/maheen-data/temp/horse_human/viz_transform_aflw_val'
visualize.writeHTMLForFolder(dir_out)
return
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
num_neighbors = 5
out_file_human = os.path.join(
out_dir_meta_face, 'matches_' + str(num_neighbors) + '_val_fiveKP.txt')
out_file_human_new = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_val_fiveKP_noIm.txt')
modifyHumanFile(out_file_human, out_file_human_new)
# out_dir_meta_face='/home/SSD3/maheen-data/horse_project/aflw';
out_file_human = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_train_fiveKP.txt')
out_file_human_new = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_train_fiveKP_noIm.txt')
modifyHumanFile(out_file_human, out_file_human_new)
return
matches_file = '/home/laoreja/data/knn_res_new/knn_5_points_val_list.txt'
face_data_file = '/home/laoreja/new-deep-landmark/dataset/train/aflw_trainImageList.txt'
# face_data_file_old='/home/laoreja/deep-landmark-master/dataset/train/trainImageList.txt';
face_data_list_file = '/home/SSD3/maheen-data/aflw_data/npy/data_list.txt'
out_dir_meta_horse = '/home/SSD3/maheen-data/horse_project/horse'
out_dir_meta_horse_list = [
os.path.join(out_dir_meta_horse, 'im'),
os.path.join(out_dir_meta_horse, 'npy')
]
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
out_dir_meta_face_list = [
os.path.join(out_dir_meta_face, 'im'),
os.path.join(out_dir_meta_face, 'npy')
]
out_dir_meta_face_old = '/home/SSD3/maheen-data/horse_project/face'
out_dir_meta_face_old_list = [
os.path.join(out_dir_meta_face_old, 'im'),
os.path.join(out_dir_meta_face_old, 'npy')
]
num_neighbors = 5
out_file_face = os.path.join(
out_dir_meta_face, 'matches_' + str(num_neighbors) + '_val_fiveKP.txt')
out_file_horse = os.path.join(
out_dir_meta_horse,
'matches_' + str(num_neighbors) + '_val_fiveKP.txt')
missing_files = makeMatchFile(num_neighbors, matches_file, face_data_file,
out_dir_meta_horse_list,
out_dir_meta_face_list, out_file_horse,
out_file_face, out_dir_meta_face_old_list)
return
matches_file = '/home/laoreja/data/knn_res_new/knn_5_points_train_list.txt'
matches_file = '/home/maheenrashid/Downloads/knn_5_points_train_list_clean.txt'
face_data_file = '/home/laoreja/new-deep-landmark/dataset/train/aflw_trainImageList.txt'
# face_data_file_old='/home/laoreja/deep-landmark-master/dataset/train/trainImageList.txt';
face_data_list_file = '/home/SSD3/maheen-data/aflw_data/npy/data_list.txt'
out_dir_meta_horse = '/home/SSD3/maheen-data/horse_project/horse'
out_dir_meta_horse_list = [
os.path.join(out_dir_meta_horse, 'im'),
os.path.join(out_dir_meta_horse, 'npy')
]
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
out_dir_meta_face_list = [
os.path.join(out_dir_meta_face, 'im'),
os.path.join(out_dir_meta_face, 'npy')
]
out_dir_meta_face_old = '/home/SSD3/maheen-data/horse_project/face'
out_dir_meta_face_old_list = [
os.path.join(out_dir_meta_face_old, 'im'),
os.path.join(out_dir_meta_face_old, 'npy')
]
num_neighbors = 5
out_file_face = os.path.join(
out_dir_meta_face,
'matches_' + str(num_neighbors) + '_train_fiveKP.txt')
out_file_horse = os.path.join(
out_dir_meta_horse,
'matches_' + str(num_neighbors) + '_train_fiveKP.txt')
missing_files = makeMatchFile(num_neighbors, matches_file, face_data_file,
out_dir_meta_horse_list,
out_dir_meta_face_list, out_file_horse,
out_file_face, out_dir_meta_face_old_list)
return
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
num_neighbors = 5
out_file_face = os.path.join(out_dir_meta_face,
'matches_' + str(num_neighbors) + '.txt')
out_file_face_new = os.path.join(
out_dir_meta_face, 'matches_noIm_' + str(num_neighbors) + '.txt')
# modifyHumanFile(out_file_face,out_file_face_new);
# old_data=util.readLinesFromFile(out_file_face);
# old_data=[line_curr.split(' ')[1] for line_curr in old_data];
# new_data=util.readLinesFromFile(out_file_face_new);
# new_data=[line_curr.split(' ')[0] for line_curr in new_data];
# assert len(old_data)==len(new_data);
# for i,old_line in enumerate(old_data):
# print i;
# assert old_line==new_data[i];
return
matches_file = '/home/laoreja/data/knn_res_new/5_points_list.txt'
matches_file = '/home/laoreja/data/knn_res_new/knn_train_list.txt'
face_data_file = '/home/laoreja/new-deep-landmark/dataset/train/aflw_trainImageList.txt'
face_data_list_file = '/home/SSD3/maheen-data/aflw_data/npy/data_list.txt'
out_dir_meta_horse = '/home/SSD3/maheen-data/horse_project/horse'
out_dir_meta_horse_list = [
os.path.join(out_dir_meta_horse, 'im'),
os.path.join(out_dir_meta_horse, 'npy')
]
out_dir_meta_face = '/home/SSD3/maheen-data/horse_project/aflw'
out_dir_meta_face_list = [
os.path.join(out_dir_meta_face, 'im'),
os.path.join(out_dir_meta_face, 'npy')
]
num_neighbors = 5
out_file_face = os.path.join(out_dir_meta_face,
'matches_' + str(num_neighbors) + '.txt')
out_file_horse = os.path.join(out_dir_meta_horse,
'matches_' + str(num_neighbors) + '.txt')
makeMatchFile(num_neighbors, matches_file, face_data_file,
out_dir_meta_horse_list, out_dir_meta_face_list,
out_file_horse, out_file_face)
return
# script_saveTrainTxt()
# dir_viz='/home/SSD3/maheen-data/temp/horse_human/viz_transform_aflw';
# visualize.writeHTMLForFolder(dir_viz,'.jpg');
return
out_dir_meta = '/home/SSD3/maheen-data'
face_dir = 'aflw_data'
horse_dir = 'horse_data'
num_neighbors = 5
path_replace_horse = [
'/home/laoreja/data/horse-images/annotation',
os.path.join(out_dir_meta, horse_dir, 'im')
]
path_replace_face = ['/npy/', '/im/']
new_match_file = os.path.join(out_dir_meta, face_dir,
'match_' + str(num_neighbors) + '.txt')
out_face_train_file = os.path.join(
out_dir_meta, face_dir, 'match_' + str(num_neighbors) + '_train.txt')
out_horse_train_file = os.path.join(
out_dir_meta, horse_dir, 'match_' + str(num_neighbors) + '_train.txt')
horse_txt_file = os.path.join(out_dir_meta, horse_dir, 'train.txt')
face_txt_file = os.path.join(out_dir_meta, face_dir, 'train.txt')
horse_train = util.readLinesFromFile(horse_txt_file)
horse_train_just_beginning = [
horse_curr.split(' ')[0] for horse_curr in horse_train
]
horse_train_just_beginning = [
horse_curr[:horse_curr.rindex('.')]
for horse_curr in horse_train_just_beginning
]
print horse_train_just_beginning[0]
face_train = util.readLinesFromFile(face_txt_file)
face_train_just_beginning = [
face_curr.split(' ')[0] for face_curr in face_train
]
face_train_just_beginning = [
face_curr[:face_curr.rindex('.')]
for face_curr in face_train_just_beginning
]
print len(horse_train)
print horse_train[0]
print len(face_train)
print face_train[0]
# return
matches = util.readLinesFromFile(new_match_file)
print(len(matches))
matches = [match_curr.split(' ') for match_curr in matches]
horse_matches = []
face_matches = []
for match_curr in matches:
assert len(match_curr) == num_neighbors + 1
horse_curr = match_curr[0]
horse_curr_path, horse_name = os.path.split(horse_curr)
if horse_curr_path[-3:] == 'gxy':
horse_curr_path = horse_curr_path[:-3]
horse_curr_path = horse_curr_path.replace(path_replace_horse[0],
path_replace_horse[1])
horse_curr = os.path.join(horse_curr_path,
horse_name[:horse_name.rindex('.')])
if horse_curr in horse_train_just_beginning:
horse_match = horse_train[horse_train_just_beginning.index(
horse_curr)]
else:
# print horse_curr
# print match_curr[0];
# raw_input();
continue
for face_curr in match_curr[1:]:
face_curr = face_curr[:face_curr.rindex('.')]
face_curr = face_curr.replace(path_replace_face[0],
path_replace_face[1])
face_match = face_train[face_train_just_beginning.index(face_curr)]
horse_matches.append(horse_match)
face_matches.append(face_match)
# print match_curr;
# print match_curr[0];
# for idx,i in enumerate(match_curr[1:]):
# print idx,face_matches[idx],i,horse_matches[idx]
assert len(face_matches) == len(horse_matches)
print len(face_matches)
util.writeFile(out_face_train_file, face_matches)
util.writeFile(out_horse_train_file, horse_matches)
return
# face_dir='/home/SSD3/maheen-data/face_data';
# train_txt=os.path.join(face_dir,'train.txt');
# files=util.readLinesFromFile(train_txt);
# files=[file_curr.split(' ') for file_curr in files];
# [im_files,npy_files]=zip(*files);
# for idx,npy_file in enumerate(npy_files):
# print idx,len(npy_files);
# assert os.path.exists(npy_file);
# assert np.load(npy_file).shape[1]==3;
# print len(im_files);
# print (im_files[0]);
# print len(npy_files);
# print (npy_files[0]);
dir_viz = '/home/SSD3/maheen-data/temp/horse_human/viz_transform'
visualize.writeHTMLForFolder(dir_viz, '.jpg')
return
horse_data = '/home/SSD3/maheen-data/horse_data'
new_face_data = '/home/SSD3/maheen-data/face_data'
old_txt = 'train.txt'
num_to_keep = 10
new_txt = 'train_' + str(num_to_keep) + '.txt'
for data_type in [horse_data, new_face_data]:
lines_new = util.readLinesFromFile(os.path.join(data_type, old_txt))
random.shuffle(lines_new)
lines_new = lines_new[:num_to_keep]
file_new = os.path.join(data_type, new_txt)
util.writeFile(file_new, lines_new)
print len(lines_new), file_new
return
def main():
train_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/train.txt'
files=util.readLinesFromFile(train_file);
random.shuffle(files);
files=files[:100];
img_paths_all=[line[:line.index(' ')] for line in files];
tif_paths_all=[line[line.index(' ')+1:] for line in files];
num_clusters=40;
out_dir='/disk3/maheen_data/debug_networks';
util.mkdir(out_dir);
out_dir_tif=os.path.join(out_dir,'tif');
util.mkdir(out_dir_tif);
out_file_html=os.path.join(out_dir,'tif_suppressCluster.html');
out_files_tif_x=[os.path.join(out_dir_tif,img_name+'_x.png') for img_name in util.getFileNames(tif_paths_all,ext='False')];
out_files_tif_y=[os.path.join(out_dir_tif,img_name+'_y.png') for img_name in util.getFileNames(tif_paths_all,ext='False')];
for tif_path,out_file_x,out_file_y in zip(tif_paths_all,out_files_tif_x,out_files_tif_y):
# print tif_path
tif=scipy.misc.imread(tif_path);
# print np.min(tif[:,:,:2]),np.max(tif[:,:,:2])
assert np.min(tif[:,:,:2])>0 and np.max(tif[:,:,:2])<num_clusters+1;
saveTifGray(tif,out_file_x,out_file_y,num_clusters)
makeImTifViz(img_paths_all,tif_paths_all,out_file_html,out_dir_tif,num_clusters=40,disk_path='disk3')
return
clusters_file='/disk2/mayExperiments/youtube_subset_new_cluster/clusters.mat';
clusters_ucf='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/clusters.mat';
min_idx_new,C_new=script_findMinCluster(clusters_file,new_flag=True);
min_idx_ucf,C_ucf=script_findMinCluster(clusters_ucf,new_flag=False);
print min_idx_new,min_idx_ucf
return
out_dir='/disk2/mayExperiments/imagenet_subset'
out_file_html=out_dir+'.html';
img_paths=util.getFilesInFolder(out_dir,'.jpg');
tif_paths=[img_path.replace('.jpg','.tif') for img_path in img_paths]
out_dir_tif=os.path.join(out_dir,'tif_viz');
util.mkdir(out_dir_tif);
makeImTifViz(img_paths,tif_paths,out_file_html,out_dir_tif);
return
train_txt='/disk2/mayExperiments/ft_youtube_hmdb_ucfClusters/train.txt';
out_dir='/disk2/mayExperiments/eval_ucf_finetune';
out_dir_tif=os.path.join(out_dir,'train_tif_select');
train_txt='/disk2/mayExperiments/ft_youtube_hmdb_newClusters/train.txt';
out_dir='/disk2/mayExperiments/eval_newClusters_finetune';
util.mkdir(out_dir);
out_dir_tif=os.path.join(out_dir,'train_tif_select');
util.mkdir(out_dir_tif);
num_to_pick=20;
num_clusters=40;
train_data=util.readLinesFromFile(train_txt);
img_paths=[line_curr[:line_curr.index(' ')] for line_curr in train_data];
tif_paths=[line_curr[line_curr.index(' ')+1:] for line_curr in train_data];
print img_paths[0].split('/');
# return
dataset=np.array([img_path.split('/')[4] for img_path in img_paths]);
print np.unique(dataset);
idx_youtube=np.where(dataset=='youtube')[0];
classes_idx=[];
classes_idx.append(np.where(dataset!='youtube')[0]);
img_paths_youtube=list(np.array(img_paths)[idx_youtube]);
img_paths_youtube_classes=np.array([img_path[:img_path.index('_')] for img_path in util.getFileNames(img_paths_youtube)])
for class_curr in np.unique(img_paths_youtube_classes):
idx_rel=np.where(img_paths_youtube_classes==class_curr)[0];
class_idx_org=idx_youtube[idx_rel];
classes_idx.append(class_idx_org);
# print len(idx_youtube);
for idx,class_idx in enumerate(classes_idx):
# print len(class_idx);
if idx>0:
paths=np.array(img_paths)[class_idx];
dataset=[img_name[:img_name.index('_')] for img_name in util.getFileNames(paths)];
# print set(dataset);
assert len(set(dataset))==1
img_paths_all=[];
tif_paths_all=[];
for class_idx in classes_idx:
img_paths_rel=np.array(img_paths)[class_idx[:num_to_pick]];
tif_paths_rel=np.array(tif_paths)[class_idx[:num_to_pick]];
img_paths_all=img_paths_all+list(img_paths_rel);
tif_paths_all=tif_paths_all+list(tif_paths_rel);
out_files_tif_x=[os.path.join(out_dir_tif,img_name+'_x.png') for img_name in util.getFileNames(tif_paths_all,ext='False')];
out_files_tif_y=[os.path.join(out_dir_tif,img_name+'_y.png') for img_name in util.getFileNames(tif_paths_all,ext='False')];
for tif_path,out_file_x,out_file_y in zip(tif_paths_all,out_files_tif_x,out_files_tif_y):
# print tif_path
tif=scipy.misc.imread(tif_path);
# print np.min(tif[:,:,:2]),np.max(tif[:,:,:2])
assert np.min(tif[:,:,:2])>0 and np.max(tif[:,:,:2])<num_clusters+1;
saveTifGray(tif,out_file_x,out_file_y,num_clusters)
out_file_html=out_dir_tif+'.html';
img_paths_html=[[util.getRelPath(img_curr) for img_curr in img_list] for img_list in zip(img_paths_all,out_files_tif_x,out_files_tif_y)];
# captions_html=[[util.getFileNames([img_curr],ext=False)[0] for img_curr in img_list] for img_list in zip(img_paths_all,out_files_tif_x,out_files_tif_y)];
captions_html=[['Image','Tif_x','Tif_y']]*len(img_paths_html);
visualize.writeHTML(out_file_html,img_paths_html,captions_html);
def main():
val_file='/disk2/mayExperiments/finetuning_youtube_hmdb_llr/val_eq.txt'
out_dir='/disk2/mayExperiments/eval_nC_zS_youtube';
model_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/opt_noFix_conv1_conv2_conv3_conv4_conv5_llr__iter_50000.caffemodel'
clusters_file='/disk3/maheen_data/youtube_train_40/clusters_100000.mat';
train_val_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/train_val_conv1_conv2_conv3_conv4_conv5.prototxt';
util.mkdir(out_dir);
out_dir_model=os.path.join(out_dir,'ft_youtube_model');
num_to_pick=100;
util.mkdir(out_dir_model);
gpu=0;
model_file_orig='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/final.caffemodel'
clusters_file_orig='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/clusters.mat';
out_dir_model_orig=os.path.join(out_dir,'ft_original_model');
util.mkdir(out_dir_model_orig);
img_paths_test=saveFlosFromValFile(out_dir_model_orig,val_file,num_to_pick,model_file_orig,clusters_file_orig,gpu,train_val_file=None,overwrite=True);
out_file_html=os.path.join(out_dir,'model_comparison.html');
out_dirs_flo_viz=[os.path.join(out_dir_model_orig,'flo_viz'),os.path.join(out_dir_model,'flo_viz')];
out_dirs_flo_viz_captions=['original_model','ft_youtube_model'];
img_paths_html=[];
captions_html=[];
img_names=util.getFileNames(img_paths_test,ext=False);
for img_path_test,img_name in zip(img_paths_test,img_names):
row_curr=[];
row_curr.append(util.getRelPath(img_path_test));
for out_dir_curr in out_dirs_flo_viz:
file_curr=os.path.join(out_dir_curr,img_name+'.png');
row_curr.append(util.getRelPath(file_curr));
captions_curr=[img_name]+out_dirs_flo_viz_captions;
img_paths_html.append(row_curr)
captions_html.append(captions_curr);
visualize.writeHTML(out_file_html,img_paths_html,captions_html);
return
path_to_npy='/disk2/mayExperiments/validation_anno/';
path_to_im='/disk2/ms_coco/val2014/';
ext='.jpg'
lim=100;
out_file=os.path.join(path_to_npy,'index_'+str(lim)+'.p');
out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz';
out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_new_3';
# out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding';
# model_file='/disk2/mayExperiments/finetuning_youtube_hmdb_llr/OptFlow_youtube_hmdb_iter_50000.caffemodel';
out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding_orig_model';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/final.caffemodel'
# out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding_ft_ucf_model';
# out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding_ft_ucf_model';
# util.mkdir(out_dir_scratch);
# model_file='/disk2/mayExperiments/ft_youtube_hmdb_ucfClusters/OptFlow_youtube_hmdb__iter_55000.caffemodel';
out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding_ft_ucf_model_imagenet';
util.mkdir(out_dir_scratch);
model_file='/disk3/maheen_data/ft_imagenet_ucf/OptFlow_imagenet_hlr__iter_22000.caffemodel'
clusters_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction_test/examples/opticalflow/clusters.mat';
out_dir_scratch='/disk2/mayExperiments/flow_resolution_scratch/im_viz_padding_ft_nC_sZ_youtube';
util.mkdir(out_dir_scratch);
model_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/opt_noFix_conv1_conv2_conv3_conv4_conv5_llr__iter_50000.caffemodel'
clusters_file='/disk3/maheen_data/youtube_train_40/clusters_100000.mat';
train_val_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/train_val_conv1_conv2_conv3_conv4_conv5.prototxt';
overwrite=True;
window=160;
step_size=16;
thresh=0.50;
scale_info=pickle.load(open(out_file,'rb'));
# scale_all=['small','large','medium'];
scale_all=['large','medium'];
scale_pow=np.arange(-2,1.5,0.5);
print scale_pow;
scale_images=[2**pow_curr for pow_curr in scale_pow];
# scale_images=scale_images[];
print scale_images;
# return
# scale_images=[0.5,1,2];
lim_cases=20;
gpu=0;
print po.NUM_THREADS
po.NUM_THREADS=multiprocessing.cpu_count();
print po.NUM_THREADS
# return
script_doEverything(path_to_npy,path_to_im,ext,lim,out_file,out_dir_scratch,window,step_size,thresh,scale_info,scale_all, scale_images,lim_cases,gpu,model_file,clusters_file,train_val_file=train_val_file,overwrite=overwrite);
script_writeHTMLForOverlap(path_to_npy,path_to_im,ext,lim,out_file,out_dir_scratch,window,step_size,thresh,scale_info,scale_all, scale_images,lim_cases,gpu,model_file,clusters_file)
def script_saveSegVizAll(params):
path_to_im_meta = params.path_to_im_meta;
path_to_im_canon = params.path_to_im_canon;
path_to_predScores_meta = params.path_to_predScores_meta;
path_to_seg_meta = params.path_to_seg_meta;
path_to_mat_overlap = params.path_to_mat_overlap;
path_to_score_mat_meta = params.path_to_score_mat_meta;
img_names_txt = params.img_names_txt;
scale_idx_range = params.scale_idx_range;
stride = params.stride;
w = params.w;
alpha = params.alpha;
num_to_pick = params.num_to_pick;
out_dir_overlay_meta = params.out_dir_overlay_meta;
power_scale_range = params.power_scale_range
power_step_size = params.power_step_size
out_file_html = params.out_file_html
height_width = params.height_width;
num_threads = params.num_threads;
out_dir_im_dict = params.out_dir_im_dict;
overwrite = params.overwrite;
util.mkdir(out_dir_overlay_meta);
for scale_curr in scale_idx_range:
util.mkdir(os.path.join(out_dir_overlay_meta,str(scale_curr)));
img_names=util.readLinesFromFile(img_names_txt);
args=[];
for idx_img_name,img_name in enumerate(img_names):
arg_curr = (img_name,path_to_im_meta,path_to_im_canon,path_to_predScores_meta,path_to_seg_meta,path_to_mat_overlap,
path_to_score_mat_meta,scale_idx_range,stride,w,alpha,num_to_pick,out_dir_overlay_meta,power_scale_range,power_step_size,out_dir_im_dict,overwrite,idx_img_name)
args.append(arg_curr);
# args=args[:10];
# args= [args[idx] for idx in [422]]
# img_list_all=[];
# for arg in args:
# img_list=saveSegVizImage(arg);
# img_list_all.append(img_list);
# for img_curr in img_list:
# print img_curr;
# break;
p=multiprocessing.Pool(num_threads);
img_list_all = p.map(saveSegVizImage,args);
img_paths_html=[];
captions_html=[];
for img_list in img_list_all:
img_row=[];
caption_row=[];
org_img=img_list[0];
# print org_img
img_list=img_list[1:];
org_img_begin = '/'.join(org_img.split('/')[:2]);
org_img = util.getRelPath(org_img,org_img_begin);
# print org_img
img_row.append(org_img);
caption_row.append(util.getFileNames([org_img])[0]);
for img_curr in img_list:
img_begin = '/'.join(img_curr.split('/')[:2]);
img_folder = img_curr.split('/')[-2];
img_row.append(util.getRelPath(img_curr,img_begin));
caption_row.append(img_folder);
img_paths_html.append(img_row);
captions_html.append(caption_row);
visualize.writeHTML(out_file_html,img_paths_html,captions_html,height=height_width[0],width=height_width[1]);
print out_file_html
def main():
# out_dir='/disk3/maheen_data/ft_youtube_40_noFix_diffLR_sZclusters';
# model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
# util.mkdir(out_dir);
# train_file=os.path.join(out_dir,'train.txt');
# template_deploy_file='trainval_noFix_withRandom_diffForConv.prototxt';
# template_solver_file='solver_debug.prototxt';
# base_lr=0.000001;
# snapshot=1000;
# layers=[None,'conv1','conv2','conv3','conv4','conv5','fc6','fc7','fc8'];
# gpu=1;
# commands=[];
# idx=len(layers)-4;
# fix_layers=layers[1:idx+1];
# layer_str='_'.join(fix_layers);
# print layer_str;
# # return
# model_file_curr=model_file
# snapshot_prefix=os.path.join(out_dir,'opt_noFix_'+layer_str+'_');
# out_deploy_file=os.path.join(out_dir,'deploy_'+layer_str+'.prototxt');
# out_solver_file=os.path.join(out_dir,'solver_'+layer_str+'.prototxt');
# log_file=os.path.join(out_dir,'log_'+layer_str+'.log');
# replaceSolverFile(out_solver_file,template_solver_file,out_deploy_file,base_lr,snapshot,snapshot_prefix,gpu=gpu);
# replaceDeployFile(out_deploy_file,template_deploy_file,train_file,fix_layers);
# command=printTrainingCommand(out_solver_file,log_file,model_file_curr);
# util.writeFile(os.path.join(out_dir,'train.sh'),[command]);
# return
model_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/opt_noFix_conv1_conv2_conv3_conv4_conv5_llr__iter_50000.caffemodel'
# model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
solver_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/solver_conv1_conv2_conv3_conv4_conv5.prototxt';
deploy_file='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/deploy_conv1_conv2_conv3_conv4_conv5.prototxt';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
solver_file='/disk3/maheen_data/ft_youtube_40_noFix_diffLR_sZclusters/solver_conv1_conv2_conv3_conv4_conv5.prototxt'
deploy_file='/disk3/maheen_data/ft_youtube_40_noFix_diffLR_sZclusters/deploy_conv1_conv2_conv3_conv4_conv5.prototxt';
justCheckGradients(solver_file,deploy_file,model_file);
return
out_dir='/disk3/maheen_data/debug_networks/sanityCheckDebug';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
util.mkdir(out_dir);
train_file=os.path.join(out_dir,'train.txt');
template_deploy_file='deploy_withRandom.prototxt';
template_solver_file='solver_debug.prototxt';
base_lr=0.000001;
snapshot=1000;
layers=[None,'conv1','conv2','conv3','conv4','conv5','fc6','fc7','fc8'];
gpu=1;
commands=[];
idx=len(layers)-1;
fix_layers=layers[1:idx+1];
layer_str='_'.join(fix_layers);
print layer_str;
model_file_curr=model_file
snapshot_prefix=os.path.join(out_dir,'opt_noFix_'+layer_str+'_');
out_deploy_file=os.path.join(out_dir,'deploy_'+layer_str+'.prototxt');
out_solver_file=os.path.join(out_dir,'solver_'+layer_str+'.prototxt');
log_file=os.path.join(out_dir,'log_'+layer_str+'.log');
replaceSolverFile(out_solver_file,template_solver_file,out_deploy_file,base_lr,snapshot,snapshot_prefix,gpu=gpu);
replaceDeployFile(out_deploy_file,template_deploy_file,train_file,fix_layers);
command=printTrainingCommand(out_solver_file,log_file,model_file_curr);
util.writeFile(os.path.join(out_dir,'train.sh'),[command]);
return
out_dir='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig/';
out_dir='/disk3/maheen_data/ft_youtube_40_images_cluster_suppress_yjConfig_llr_diff/';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
util.mkdir(out_dir);
train_file=os.path.join(out_dir,'train.txt');
template_deploy_file='deploy_withRandom_yjConfig.prototxt';
template_solver_file='solver_debug.prototxt';
base_lr=0.00001;
snapshot=500;
layers=[None,'conv1','conv2','conv3','conv4','conv5','fc6','fc7','fc8'];
gpu=1;
commands=[];
idx=len(layers)-4;
fix_layers=layers[1:idx+1];
layer_str='_'.join(fix_layers);
print layer_str;
# return
model_file_curr=model_file
snapshot_prefix=os.path.join(out_dir,'opt_noFix_'+layer_str+'_');
out_deploy_file=os.path.join(out_dir,'deploy_'+layer_str+'.prototxt');
out_solver_file=os.path.join(out_dir,'solver_'+layer_str+'.prototxt');
log_file=os.path.join(out_dir,'log_'+layer_str+'.log');
replaceSolverFile(out_solver_file,template_solver_file,out_deploy_file,base_lr,snapshot,snapshot_prefix,gpu=gpu);
replaceDeployFile(out_deploy_file,template_deploy_file,train_file,fix_layers);
command=printTrainingCommand(out_solver_file,log_file,model_file_curr);
util.writeFile(os.path.join(out_dir,'train.sh'),[command]);
return
out_dir='/disk3/maheen_data/ft_youtube_40_noFix_noCopyFC8_FC7';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
# out_dir='/disk3/maheen_data/ft_youtube_40_noFix_alexnet';
# model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/models/bvlc_alexnet/bvlc_alexnet.caffemodel';
util.mkdir(out_dir);
train_txt_orig_path='/disk3/maheen_data/ft_youtube_40/train.txt';
template_deploy_file='deploy_withRandom.prototxt';
template_solver_file='solver_debug.prototxt';
train_file=os.path.join(out_dir,'train.txt');
data=util.readLinesFromFile(train_txt_orig_path);
random.shuffle(data);
# data[:100];
util.writeFile(train_file,data);
# shutil.copyfile(train_txt_orig_path,train_file);
# out_dir='/disk3/maheen_data/ft_youtube_40_ucf_permute';
# train_file=os.path.join(out_dir,'train_permute.txt');
base_lr=0.0001;
snapshot=2000;
layers=[None,'conv1','conv2','conv3','conv4','conv5','fc6','fc7','fc8'];
gpu=0;
# command_file=os.path.join(out_dir,'debug_0.sh');
commands=[];
# for idx in range(4,len(layers)):
# if idx==0:
# fix_layers=layers[0];
# layer_str=str(fix_layers);
# model_file_curr=None;
# else:
idx=len(layers)-3;
fix_layers=layers[1:idx+1];
layer_str='_'.join(fix_layers);
print layer_str;
return
model_file_curr=model_file
# print fix_layers
snapshot_prefix=os.path.join(out_dir,'opt_noFix_'+layer_str+'_');
out_deploy_file=os.path.join(out_dir,'deploy_'+layer_str+'.prototxt');
out_solver_file=os.path.join(out_dir,'solver_'+layer_str+'.prototxt');
log_file=os.path.join(out_dir,'log_'+layer_str+'.log');
replaceSolverFile(out_solver_file,template_solver_file,out_deploy_file,base_lr,snapshot,snapshot_prefix,gpu=gpu);
replaceDeployFile(out_deploy_file,template_deploy_file,train_file,fix_layers);
command=printTrainingCommand(out_solver_file,log_file,model_file_curr);
util.writeFile(os.path.join(out_dir,'train.sh'),[command]);
# commands.append(command);
# util.writeFile(command_file,commands);
return
# out_dir='/disk3/maheen_data/debug_networks/noFix';
# model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
# '/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/train.txt'
# data=util.readLinesFromFile(train_txt_orig_path);
# random.shuffle(data);
# # data[:100];
# util.writeFile(train_file,data[:100]);
# out_dir='/disk3/maheen_data/debug_networks/noFixNoCopy';
# model_file=None;
out_dir='/disk3/maheen_data/debug_networks/noFixCopyByLayer';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
util.mkdir(out_dir);
train_txt_orig_path='/disk3/maheen_data/debug_networks/noFix/train.txt';
deploy_file='/disk3/maheen_data/debug_networks/noFix/deploy.prototxt';
solver_file='/disk3/maheen_data/debug_networks/noFix/solver.prototxt';
# template_deploy_file='deploy_debug_noFix.prototxt';
template_deploy_file='deploy_fc8NoCopy.prototxt';
template_solver_file='solver_debug.prototxt';
train_file=os.path.join(out_dir,'train.txt');
# shutil.copyfile(train_txt_orig_path,train_file);
base_lr=0.0001;
snapshot=100;
layers=[None,'conv1','conv2','conv3','conv4','conv5','fc6','fc7','fc8'];
return
out_dir='/disk3/maheen_data/ft_youtube_40_noFix_alexnet';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/models/bvlc_alexnet/bvlc_alexnet.caffemodel';
layers=['conv1','conv2','conv3','conv4','conv5','fc6','fc7'];
layers_str=[];
for idx in range(len(layers)):
if idx==0:
fix_layers=layers[0];
layer_str=str(fix_layers);
else:
fix_layers=layers[1:idx+1];
layer_str='_'.join(fix_layers);
layers_str.append(layer_str);
log_files=[os.path.join(out_dir,'log_'+layer_str+'.log') for layer_str in layers_str];
str_match=' solver.cpp:209] Iteration ';
xAndYs=[svl.getIterationsAndLosses(log_file,str_match) for log_file in log_files];
out_files=[];
for layer_str,log_file in zip(layers_str,log_files):
xAndY=svl.getIterationsAndLosses(log_file,str_match);
print xAndY
out_file=os.path.join(out_dir,'loss_'+layer_str+'.png');
visualize.plotSimple([xAndY],out_file,title=layer_str);
out_files.append(out_file);
out_file_html=os.path.join(out_dir,'losses_all.html');
img_paths=[[util.getRelPath(out_file,'/disk3')] for out_file in out_files];
captions=[['']]*len(out_files);
print img_paths
print captions
visualize.writeHTML(out_file_html,img_paths,captions,height=300,width=300);
# out_file=os.path.join(out_dir,'losses_all.png');
# print len(xAndYs);
# print xAndYs[-2][1]
# visualize.plotSimple(xAndYs,out_file,legend_entries=layers_str,loc=0,outside=True)
return
# mean_standard_proto_file='/home/maheenrashid/Downloads/debugging_jacob/opticalflow/standard.binaryproto';
model_file='/home/maheenrashid/Downloads/debugging_jacob/optical_flow_prediction/examples/opticalflow/final.caffemodel';
layers_to_copy=['conv1','conv2','conv3','conv4','conv5']
# model_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/OptFlow_youtube_hmdb__iter_5000.caffemodel';
# layers_to_copy=['conv1','conv2','conv3','conv4','conv5','fc6_fix','fc7_fix','fc8_fix']
# model_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic_llr/OptFlow_youtube_hmdb__iter_65000.caffemodel';
# layers_to_copy=['conv1','conv2','conv3','conv4','conv5','fc6_fix','fc7_fix','fc8_fix']
# deploy_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/opt_train_coarse_xavier.prototxt';
# solver_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/train.prototxt';
deploy_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/deploy_debug.prototxt';
solver_file='/disk2/mayExperiments/ft_youtube_hmdb_newClusters_layerMagic/solver_debug.prototxt';
# layers_to_copy=['conv1','conv2','conv3','conv4','conv5','fc6_fix','fc7_fix','fc8_fix']
# layers_to_explore=['conv1','conv2','conv3','conv4','conv5','fc6_fix','fc7_fix','fc8_fix']
# ,'fc6','fc7','fc8']
layers_to_explore=['conv1','conv2','conv3','conv4','conv5','fc6_fix','fc7_fix','fc8_fix']
blobs_lr=[(0,0),(0,0),(0,0),
# (10,20),(10,20),
(0.1,0.2),(0.1,0.2),
(1,2),(1,2),(1,2)]
iterations=400;
momentum=0.9;
lr=0.000001;
caffe.set_device(1)
caffe.set_mode_gpu()
solver=caffe.SGDSolver(solver_file);
net_org=caffe.Net(deploy_file,model_file);
# copy weights
for layer_name in layers_to_copy:
solver.net.params[layer_name][0].data[...]=net_org.params[layer_name][0].data;
solver.net.params[layer_name][1].data[...]=net_org.params[layer_name][1].data;
layer_names=list(solver.net._layer_names);
ratios={};
for key in layers_to_explore:
ratios[key]=[];
dict_layers={};
for idx_curr,layer_name in enumerate(layer_names):
print idx_curr,layer_name,
if layer_name in solver.net.params.keys():
print len(solver.net.params[layer_name])
update_prev=[np.zeros(solver.net.layers[idx_curr].blobs[0].diff.shape),
np.zeros(solver.net.layers[idx_curr].blobs[1].diff.shape)];
blob_lr=list(blobs_lr[layers_to_explore.index(layer_name)]);
dict_layers[idx_curr]=[layer_name,update_prev,blob_lr];
else:
print 0;
for idx_curr in dict_layers.keys():
print idx_curr,len(dict_layers[idx_curr]),dict_layers[idx_curr][0],dict_layers[idx_curr][1][0].shape,dict_layers[idx_curr][1][1].shape,dict_layers[idx_curr][2]
for iteration in range(iterations):
print iteration
solver.net.forward();
solver.net.backward();
for idx_curr in dict_layers.keys():
rel_row=dict_layers[idx_curr]
layer_name=rel_row[0];
update_prev=rel_row[1][0];
print rel_row[2][0]
lr_curr=rel_row[2][0]*lr;
diffs_curr=solver.net.params[layer_name][0].diff;
weights_curr=solver.net.params[layer_name][0].data;
param_scale = np.linalg.norm(weights_curr.ravel())
update = update_prev*momentum-lr_curr*diffs_curr;
update_scale = np.linalg.norm(update.ravel())
ratio= update_scale / param_scale # want ~1e-3
print layer_name,ratio,update_scale,param_scale
ratios[layer_name].append(ratio);
for idx_curr,layer in enumerate(solver.net.layers):
for idx_blob,blob in enumerate(layer.blobs):
rel_row=dict_layers[idx_curr]
layer_name=rel_row[0];
update_prev=rel_row[1][idx_blob];
lr_curr=rel_row[2][idx_blob]*lr;
diffs_curr=blob.diff;
update_curr=momentum*update_prev-(lr_curr*diffs_curr);
blob.data[...] -= update_curr
blob.diff[...] = np.zeros(blob.diff.shape);
dict_layers[idx_curr][1][idx_blob]=update_curr;
def main():
old_file = '/home/laoreja/data/knn_res_new/knn_5_points_train_list.txt'
new_file = '/home/maheenrashid/Downloads/knn_5_points_train_list_clean.txt'
match_str = 'n02374451_4338.JPEG'
lines = util.readLinesFromFile(old_file)
lines_to_keep = []
for line in lines:
if match_str not in line:
lines_to_keep.append(line)
assert len(lines_to_keep) == len(lines) - 1
util.writeFile(new_file, lines_to_keep)
return
file_curr = '/home/laoreja/finetune-deep-landmark/dataset/train/trainImageList_2.txt'
out_file = '/home/maheenrashid/Downloads/trainImageList_2_clean.txt'
lines = util.readLinesFromFile(file_curr)
lines_to_keep = []
for line in lines:
if line == '/home/laoreja/data/horse-images/annotation/imagenet_n02374451/gxy/n02374451_4338.JPEG 156 169 79 99 161 88 1 43 46 1 167 95 1 164 95 1 43 56 1':
print 'found!'
else:
lines_to_keep.append(line)
print len(lines_to_keep), len(lines)
assert len(lines_to_keep) + 1 == len(lines)
util.writeFile(out_file, lines_to_keep)
return
horse_file = '/home/SSD3/maheen-data/horse_project/horse/matches_5_val_fiveKP.txt'
human_file = '/home/SSD3/maheen-data/horse_project/aflw/matches_5_val_fiveKP_noIm.txt'
horse_data = util.readLinesFromFile(horse_file)
human_data = util.readLinesFromFile(human_file)
# horse_data=[horse_data[41]];
# human_data=[human_data[41]];
# print horse_data[0];
horse_im = [line_curr.split(' ')[0] for line_curr in horse_data]
human_im = [
line_curr.split(' ')[0].replace('/npy/',
'/im/').replace('.npy', '.jpg')
for line_curr in human_data
]
horse_npy = [line_curr.split(' ')[1] for line_curr in horse_data]
human_npy = [line_curr.split(' ')[0] for line_curr in human_data]
problem_cases = []
for horse_npy_curr in horse_npy:
labels = np.load(horse_npy_curr)
if np.any(labels < 0):
problem_cases.append(horse_npy_curr)
print len(problem_cases), len(set(problem_cases))
return
dir_server = '/home/SSD3/maheen-data'
out_dir_debug = os.path.join(dir_server, 'temp',
'debug_problem_batch/rerun')
im_file = '/home/laoreja/data/horse-images/annotation/imagenet_n02374451/gxy/n02374451_4338.JPEG'
npy_file = '/home/SSD3/maheen-data/temp/debug_problem_batch/rerun/npy/imagenet_n02374451/n02374451_4338.npy'
out_file = os.path.join(out_dir_debug, 'check.png')
saveImWithAnno((1, im_file, npy_file, out_file))
# arg=([156, 169, 79, 99], [[161, 88, 1], [43, 46, 1], [167, 95, 1], [164, 95, 1], [43, 56, 1]], '/home/SSD3/maheen-data/temp/debug_problem_batch/rerun/npy/imagenet_n02374451/n02374451_4338.npy', 0);
# # print np.load(arg[2]);
# saveBBoxNpy(arg);
# # print np.load(arg[2]);
return
dir_server = '/home/SSD3/maheen-data'
out_dir_debug = os.path.join(dir_server, 'temp',
'debug_problem_batch/rerun')
util.mkdir(out_dir_debug)
params_dict = {}
params_dict[
'path_txt'] = '/home/SSD3/maheen-data/temp/debug_problem_batch/train_dummy.txt'
# '/home/laoreja/finetune-deep-landmark/dataset/train/trainImageList_2.txt';
params_dict['path_pre'] = None
params_dict['type_data'] = 'horse'
params_dict['out_dir_meta'] = out_dir_debug
util.mkdir(params_dict['out_dir_meta'])
params_dict['out_dir_im'] = os.path.join(params_dict['out_dir_meta'], 'im')
params_dict['out_dir_npy'] = os.path.join(params_dict['out_dir_meta'],
'npy')
params_dict['out_file_list_npy'] = os.path.join(params_dict['out_dir_npy'],
'data_list.txt')
params_dict['out_file_list_im'] = os.path.join(params_dict['out_dir_im'],
'data_list.txt')
params_dict['out_file_pairs'] = os.path.join(params_dict['out_dir_meta'],
'pairs.txt')
params_dict['overwrite'] = True
local_script_makeBboxPairFiles(params_dict)
return
npy_file = '/home/SSD3/maheen-data/horse_project/horse/npy/imagenet_n02374451/n02374451_4338.npy'
labels = np.load(npy_file)
print labels
return
dir_server = '/home/SSD3/maheen-data'
out_dir_debug = os.path.join(dir_server, 'temp', 'debug_problem_batch')
util.mkdir(out_dir_debug)
out_horse_im_dir = os.path.join(out_dir_debug, 'horse_im')
out_human_im_dir = os.path.join(out_dir_debug, 'human_im')
util.mkdir(out_horse_im_dir)
util.mkdir(out_human_im_dir)
horse_file = '/home/SSD3/maheen-data/horse_project/horse_resize/matches_5_train_fiveKP_debug.txt'
human_file = '/home/SSD3/maheen-data/horse_project/aflw/matches_5_train_fiveKP_noIm_debug.txt'
horse_data = util.readLinesFromFile(horse_file)
human_data = util.readLinesFromFile(human_file)
horse_data = [horse_data[41]]
human_data = [human_data[41]]
print horse_data[0]
horse_im = [line_curr.split(' ')[0] for line_curr in horse_data]
human_im = [
line_curr.split(' ')[0].replace('/npy/',
'/im/').replace('.npy', '.jpg')
for line_curr in human_data
]
horse_npy = [line_curr.split(' ')[1] for line_curr in horse_data]
human_npy = [line_curr.split(' ')[0] for line_curr in human_data]
args = []
for idx, horse_im_curr in enumerate(horse_im):
args.append((idx, horse_im_curr, horse_npy[idx],
os.path.join(out_horse_im_dir,
str(idx) + '.jpg')))
for idx, horse_im_curr in enumerate(human_im):
args.append((idx, horse_im_curr, human_npy[idx],
os.path.join(out_human_im_dir,
str(idx) + '.jpg')))
# saveImWithAnno(args[-1]);
p = multiprocessing.Pool(multiprocessing.cpu_count())
p.map(saveImWithAnno, args)
out_file_html = os.path.join(out_dir_debug, 'viz_matches.html')
img_paths = []
captions = []
for idx in range(len(horse_im)):
horse_im_curr = os.path.join(out_horse_im_dir,
str(idx) + '.jpg')
horse_im_curr = util.getRelPath(horse_im_curr, dir_server)
human_im_curr = os.path.join(out_human_im_dir,
str(idx) + '.jpg')
human_im_curr = util.getRelPath(human_im_curr, dir_server)
img_paths.append([horse_im_curr, human_im_curr])
captions.append(['horse ' + str(idx), 'human'])
# for idx,horse_im_curr in enumerate(horse_im):
# human_im_curr=util.getRelPath(human_im[idx],dir_server);
# horse_im_curr=util.getRelPath(horse_im_curr,dir_server);
# img_paths.append([horse_im_curr,human_im_curr]);
# captions.append(['horse','human']);
visualize.writeHTML(out_file_html, img_paths, captions, 224, 224)
