def script_make_im_gray():
dir_meta = '../data/disfa'
out_dir_im = os.path.join(dir_meta, 'Videos_LeftCamera_frames_200')
# ../data/disfa/preprocess_im_200_color_align
# out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color_align')
# out_dir_files_new = os.path.join(dir_meta, 'train_test_files_110_gray_align')
out_dir_im_new = os.path.join(dir_meta, 'preprocess_im_110_gray_align')
util.mkdir(out_dir_im_new)
num_folds = 3
im_size = [110, 110]
# [96,96]
# all_im = []
# for fold_curr in range(num_folds):
# train_file = os.path.join(out_dir_files,'train_'+str(fold_curr)+'.txt')
# test_file = os.path.join(out_dir_files,'test_'+str(fold_curr)+'.txt')
# all_data = util.readLinesFromFile(train_file)+util.readLinesFromFile(test_file)
# all_im = all_im + [line_curr.split(' ')[0] for line_curr in all_data]
all_im = glob.glob(os.path.join(out_dir_im, '*', '*.jpg'))
print len(all_im), len(set(all_im))
all_im = list(set(all_im))
args = []
for idx_file_curr, file_curr in enumerate(all_im):
out_file_curr = file_curr.replace(out_dir_im, out_dir_im_new)
dir_curr = os.path.split(out_file_curr)[0]
util.makedirs(dir_curr)
# print out_file_curr
# print dir_curr
if not os.path.exists(out_file_curr):
args.append((file_curr, out_file_curr, im_size, idx_file_curr))
print len(args)
# for arg in args:
# print arg
# preprocess_bp4d.save_color_as_gray(arg)
# raw_input()
pool = multiprocessing.Pool(multiprocessing.cpu_count())
pool.map(preprocess_bp4d.save_color_as_gray, args)
def make_rough_anno_mat(test = False):
dir_meta_curr = os.path.join(dir_meta, 'val_data')
if test:
dir_meta_curr = os.path.join(dir_meta, 'test_data')
anno_dir = os.path.join(dir_meta_curr, 'annotation')
out_dir_anno_rough = os.path.join(dir_meta_curr, 'anno_rough')
util.mkdir(out_dir_anno_rough)
files_anno = glob.glob(os.path.join(anno_dir,'*.txt'))
files_anno = [file_curr for file_curr in files_anno if not os.path.split(file_curr)[1].startswith('Ambiguous')]
classes = [os.path.split(file_curr[:file_curr.rindex('_')])[1] for file_curr in files_anno]
classes.sort()
assert len(classes)==20
dict_annos = {}
for file_curr in files_anno:
annos = util.readLinesFromFile(file_curr)
class_curr = os.path.split(file_curr[:file_curr.rindex('_')])[1]
anno_num = classes.index(class_curr)
for anno_curr in annos:
anno_curr = anno_curr.split()
key = anno_curr[0]
start = float(anno_curr[1])
end = float(anno_curr[2])
if key not in dict_annos:
dict_annos[key] = [[] for idx in range(20)]
dict_annos[key][anno_num].append([start, end])
for key in dict_annos.keys():
annos = dict_annos[key]
annos = [np.array(anno_curr) for anno_curr in annos]
assert len(annos)==20
out_file = os.path.join(out_dir_anno_rough, key+'.npy')
# print out_file
np.save(out_file, annos)
def script_save_bbox():
dir_meta = '../data/bp4d'
in_dir_meta = os.path.join(dir_meta,
'preprocess_im_' + str(256) + '_color_nodetect')
out_dir_meta = os.path.join(
dir_meta, 'preprocess_im_' + str(256) + '_color_nodetect_bbox')
util.mkdir(out_dir_meta)
im_list_in = glob.glob(os.path.join(in_dir_meta, '*', '*', '*.jpg'))
print len(im_list_in)
savegray = False
args = []
args = []
file_pairs = []
for idx_im_in, im_in in enumerate(im_list_in):
out_file = im_in.replace(in_dir_meta,
out_dir_meta).replace('.jpg', '.npy')
if os.path.exists(out_file):
continue
out_dir_curr = os.path.split(out_file)[0]
util.makedirs(out_dir_curr)
file_pairs.append((im_in, out_file))
chunk_size = 5
chunks = [
file_pairs[x:x + chunk_size]
for x in range(0, len(file_pairs), chunk_size)
]
args = [(chunk_curr, idx_im_in)
for idx_im_in, chunk_curr in enumerate(chunks)]
print len(args)
# args = args[:1000]
for arg in args:
print arg
size = save_best_bbox_batch(arg)
raw_input()
def write_train_test_au():
dir_meta = '../data/pain'
au_dir = os.path.join(dir_meta, 'anno_au')
facs_dir = os.path.join(dir_meta, 'Frame_Labels','FACS')
im_dir = os.path.join(dir_meta, 'preprocess_256_color')
gt_pain_dir = os.path.join(dir_meta, 'Sequence_Labels', 'gt_avg')
train_test_dir = os.path.join(dir_meta,'train_test_files_loo_1_thresh_au_only')
util.mkdir(train_test_dir)
strs_replace = [[facs_dir,im_dir],['_facs.txt','.jpg']]
people_dirs = [os.path.split(dir_curr)[1] for dir_curr in glob.glob(os.path.join(au_dir,'*'))]
people_dirs.sort()
for idx_test_dir, test_dir in enumerate(people_dirs):
train_file = os.path.join(train_test_dir,'train_'+str(idx_test_dir)+'.txt')
test_file = os.path.join(train_test_dir,'test_'+str(idx_test_dir)+'.txt')
train_lines = []
num_skipped_all = 0
for idx_people_dir, people_dir in enumerate(people_dirs):
if idx_people_dir!=idx_test_dir:
# train_dirs = [dir_curr for dir_curr in people_dirs if dir_curr!=test_dir]
train_lines_curr, num_skipped = get_threshold_pain_au_only(au_dir,people_dir,gt_pain_dir, strs_replace)
num_skipped_all += num_skipped
train_lines += train_lines_curr
else:
test_lines, _ = get_threshold_pain_au_only(au_dir,people_dir,gt_pain_dir, strs_replace,threshold=0)
print num_skipped_all
print train_file, len(train_lines)
print test_file, len(test_lines)
# random.shuffle(train_lines)
# random.shuffle(test_lines)
util.writeFile(train_file, train_lines)
util.writeFile(test_file, test_lines)
def create_256_train_test_files():
in_data_dir = '../data/ck_96/train_test_files'
out_data_dir = '../data/ck_256/train_test_files'
util.mkdir(out_data_dir)
num_folds = 10
for split_num in range(0, num_folds):
for file_pre in ['train', 'test']:
in_file = os.path.join(in_data_dir,
file_pre + '_' + str(split_num) + '.txt')
out_file = os.path.join(out_data_dir,
file_pre + '_' + str(split_num) + '.txt')
in_lines = util.readLinesFromFile(in_file)
# print in_lines[0]
# raw_input()
out_lines = [
line_curr.replace(in_data_dir, out_data_dir)
for line_curr in in_lines
]
print out_file
util.writeFile(out_file, out_lines)
def make_train_test_files():
dir_meta = '../data/bp4d'
out_dir_subs = os.path.join(dir_meta, 'subs')
out_dir_annos = os.path.join(dir_meta, 'anno_text')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_color_nodetect')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color_nodetect')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_color')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_96_gray')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_96_gray')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_color_align')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color_align')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_gray_align')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_110_gray_align')
# out_dir_im = os.path.join(dir_meta, 'preprocess_im_256_color_align')
# out_dir_files = os.path.join(dir_meta, 'train_test_files_256_color_align')
out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_gray_nodetect')
out_dir_files = os.path.join(dir_meta,
'train_test_files_110_gray_nodetect')
replace_str = '../data/bp4d/BP4D/BP4D-training'
util.mkdir(out_dir_files)
num_folds = 3
for fold_num in range(num_folds):
for file_pre_str in ['train', 'test']:
train_sub_file = os.path.join(
out_dir_subs, file_pre_str + '_' + str(fold_num) + '.txt')
train_folds = util.readLinesFromFile(train_sub_file)
out_file_train = os.path.join(
out_dir_files, file_pre_str + '_' + str(fold_num) + '.txt')
write_train_file(out_file_train, out_dir_annos, out_dir_im,
train_folds, replace_str)
def split_and_save_i3d():
np_file ='../data/i3d_features/Thumos14-I3D-JOINTFeatures.npy'
videos_path = '../data/ucf101/val_data/validation'
out_dir = '../data/i3d_features/Thumos14-I3D-JOINTFeatures_val'
test=False
util.mkdir(out_dir)
videos_path = '../data/ucf101/test_data/TH14_test_set_mp4'
out_dir = '../data/i3d_features/Thumos14-I3D-JOINTFeatures_test'
test=True
util.mkdir(out_dir)
val_videos = glob.glob(os.path.join(videos_path, '*.mp4'))
val_videos.sort()
print len(val_videos)
print val_videos[0]
print val_videos[-1]
features = np.load(np_file)
if test:
features = features[-len(val_videos):]
else:
features = features[:len(val_videos)]
print len(features),len(val_videos)
for features_idx, features_curr in enumerate(features[:len(val_videos)]):
# print features_idx, val_videos[features_idx]
# if features_idx<1009:
# continue
out_file = os.path.join(out_dir,os.path.split(val_videos[features_idx])[1].replace('.mp4','.npy'))
# print out_file
# print features_curr.shape
print features_idx,out_file,features_curr.shape
# raw_input()
np.save(out_file,features_curr)
def save_sim_viz(vid_name,
out_shape_curr,
sim_mat,
class_idx,
out_dir,
dataset='ucf'):
gt_vals, det_times = get_gt_vector(vid_name,
out_shape_curr,
class_idx,
dataset=dataset)
if dataset.startswith('activitynet'):
class_names = globals.class_names_activitynet
else:
class_names = globals.class_names
out_dir_curr = os.path.join(out_dir, class_names[class_idx])
util.mkdir(out_dir_curr)
pos_rows = sim_mat[gt_vals > 0, :]
pos_rows = np.mean(pos_rows, axis=0)
neg_rows = sim_mat[gt_vals < 1, :]
neg_rows = np.mean(neg_rows, axis=0)
# for idx_pos_row, pos_row in enumerate(pos_rows):
max_val = max(np.max(pos_rows), np.max(neg_rows))
gt_vals_curr = gt_vals * max_val
arr_plot = [(det_times, curr_arr)
for curr_arr in [gt_vals_curr, pos_rows, neg_rows]]
legend_entries = ['gt', 'pos', 'neg']
# idx_pos_row = str(idx_pos_row)
out_file_curr = os.path.join(out_dir_curr, vid_name + '.jpg')
title = vid_name
# +' '+idx_pos_row
# visualize.plotSimple(arr_plot, out_file = out_file_curr, title = title, xlabel = 'time', ylabel = 'max sim', legend_entries = legend_entries)
# print out_file_curr
# print
np.save(out_file_curr.replace('.jpg', '.npy'), sim_mat)
visualize.saveMatAsImage(sim_mat, out_file_curr, title=title)
def create_comparative_html_fast():
just_train = True
# load files
dir_train_test_files = '../data/ucf101/train_test_files'
train_file = os.path.join(dir_train_test_files,
'train_just_primary_corrected.txt')
test_file = os.path.join(dir_train_test_files,
'test_just_primary_corrected.txt')
dir_gt_vecs = '../data/ucf101/gt_vecs/just_primary_corrected'
if just_train:
out_dir = '../scratch/i3d_dists_just_train'
out_dir_coocs = '../scratch/i3d_dists_just_train/arr_coocs_pairwise'
out_dir_viz = '../scratch/i3d_dists_just_train/arr_coocs_pairwise_viz'
util.mkdir(out_dir_viz)
lines = util.readLinesFromFile(train_file)
else:
out_dir = '../scratch/i3d_dists'
lines = util.readLinesFromFile(train_file) + util.readLinesFromFile(
test_file)
npy_files = [line_curr.split(' ')[0] for line_curr in lines]
gt_files = [
os.path.join(dir_gt_vecs,
os.path.split(line_curr)[1]) for line_curr in npy_files
]
# get vid names per class
vid_names_per_class, class_id = get_vid_names_per_class(lines)
# print vid_names_per_class
out_dir_html = out_dir + '_pairwise_htmls'
util.mkdir(out_dir_html)
dir_server = '/disk2/maheen_data'
out_dir_viz = out_dir_viz.replace('..', os.path.join(dir_server, 'nn_net'))
for dir_type in ['fg', 'bg', 'mat']:
out_dir_html_curr = os.path.join(out_dir_html, dir_type)
util.mkdir(out_dir_html_curr)
n_strs = ['n_10', 'n_25', 'n_50', 'n_100']
folders = [
os.path.join(out_dir_viz, n_str, dir_type) for n_str in n_strs
]
for idx_class, vid_names_curr in enumerate(vid_names_per_class):
class_name = class_names[idx_class]
img_names = [class_name + '.jpg'] + [
vid_name + '.jpg' for vid_name in vid_names_curr
]
out_file_html = os.path.join(out_dir_html_curr,
class_name + '.html')
visualize.writeHTMLForDifferentFolders(out_file_html,
folders,
n_strs,
img_names,
rel_path_replace=dir_server,
height=330,
width=400)
def script_median_of_all():
# frame_folder = '../data/intervals_for_extraction_128_128_0.1fps/'
# out_folder = '../data/median_bg'
# skip_num = 1
frame_folder = '../data/pain_no_pain_x2h_intervals_for_extraction_672_380_0.2fps/'
out_folder = '../data/testing_median'
skip_num = 100
util.mkdir(out_folder)
path_list = get_all_jpg_paths(frame_folder)
path_list = path_list[::skip_num]
per_cam_lists = []
# Make list of empty lists, one per camera
for i in range(NUM_CAMERAS):
per_cam_list = []
per_cam_lists.append(per_cam_list)
# Read and sort images into the right camera list
with tqdm(total=len(path_list)) as pbar:
for idx, path in enumerate(path_list):
pbar.update(1)
camera = int(get_camera_from_path(path))
cam_idx = camera - 1 # Make zero-indexed
img = imread(path)
per_cam_lists[cam_idx].append(img)
# Per camera: convert to array, compute median and save
with tqdm(total=NUM_CAMERAS) as pbar:
for i in range(NUM_CAMERAS):
pbar.update(1)
ar = np.asarray(per_cam_lists[i])
med = np.median(ar, axis=0)
imsave(
os.path.join(out_folder,
'median_0.1fps_camera_{}.jpg'.format(i)),
med.astype('uint8'))
def script_make_im_gray():
dir_meta = '../data/bp4d'
out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_color_align')
out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color_align')
out_dir_files_new = os.path.join(dir_meta, 'train_test_files_110_gray_align')
out_dir_im_new = os.path.join(dir_meta, 'preprocess_im_110_gray_align')
util.mkdir(out_dir_files_new)
num_folds = 3
im_size = None
# [96,96]
all_im = []
for fold_curr in range(num_folds):
train_file = os.path.join(out_dir_files,'train_'+str(fold_curr)+'.txt')
test_file = os.path.join(out_dir_files,'test_'+str(fold_curr)+'.txt')
all_data = util.readLinesFromFile(train_file)+util.readLinesFromFile(test_file)
all_im = all_im + [line_curr.split(' ')[0] for line_curr in all_data]
print len(all_im ),len(set(all_im))
all_im = list(set(all_im))
args = []
for idx_file_curr,file_curr in enumerate(all_im):
out_file_curr = file_curr.replace(out_dir_im,out_dir_im_new)
dir_curr = os.path.split(out_file_curr)[0]
util.makedirs(dir_curr)
# print out_file_curr
# print dir_curr
if not os.path.exists(out_file_curr):
args.append((file_curr,out_file_curr,im_size,idx_file_curr))
print len(args)
# for arg in args:
# print arg
# save_color_as_gray(arg)
# raw_input()
pool = multiprocessing.Pool(multiprocessing.cpu_count())
pool.map(save_color_as_gray,args)
def save_npys():
meta_dir = '../data/charades'
out_dir = os.path.join(meta_dir, 'vgg16_rgb_features_npy')
util.mkdir(out_dir)
train_file = os.path.join(meta_dir, 'train_test_files',
'vgg_16_rgb_test.txt')
lines = util.readLinesFromFile(train_file)
# t = time.time()
for idx_line_curr, line_curr in enumerate(lines):
if idx_line_curr % 100 == 0:
print idx_line_curr, len(lines)
file_curr = line_curr.split(' ')[0]
out_file = os.path.join(out_dir,
os.path.split(file_curr)[1]).replace(
'.npz', '.npy')
if not os.path.exists(out_file):
feat = np.load(file_curr)['arr_0']
np.save(out_file, feat)
def save_kp_for_alignment():
dir_meta = '../data/bp4d'
out_dir_im = os.path.join(dir_meta, 'preprocess_im_110_color')
out_dir_files = os.path.join(dir_meta, 'train_test_files_110_color')
out_dir_kp = os.path.join(dir_meta,'preprocess_im_110_color_kp')
util.mkdir(out_dir_kp)
all_im = []
for fold_num in range(3):
for file_pre in ['train','test']:
file_curr = os.path.join(out_dir_files,file_pre+'_'+str(fold_num)+'.txt')
im_list_curr = [line.split(' ')[0] for line in util.readLinesFromFile(file_curr)]
all_im.extend(im_list_curr)
all_im = list(set(all_im))
print len(all_im)
batch_size = 128
batches = [all_im[x:x+batch_size] for x in range(0, len(all_im), batch_size)]
total = 0
for b in batches:
total +=len(b)
assert total==len(all_im)
fa=face_alignment.FaceAlignment(face_alignment.LandmarksType._2D, enable_cuda=True, flip_input=False)
for idx_im_list, im_list in enumerate(batches):
print 'batch',idx_im_list,'of',len(batches)
preds = fa.get_landmarks_simple(im_list)
# print preds.shape,len(im_list)
for idx_im_curr,im_curr in enumerate(im_list):
pred_curr = preds[idx_im_curr]
# print pred_curr.shape
out_file_pred = im_curr.replace(out_dir_im,out_dir_kp).replace('.jpg','.npy')
# print out_file_pred,im_curr
dir_curr = os.path.split(out_file_pred)[0]
util.makedirs(dir_curr)
# raw_input()
np.save(out_file_pred,pred_curr)
def save_flows_gpu((in_dir_meta, out_dir_u, out_dir_v, video_name, sample,
idx)):
print idx
in_dir_curr = os.path.join(in_dir_meta, video_name)
in_files = glob.glob(os.path.join(in_dir_curr, '*.jpg'))
in_files.sort()
# print len(in_files)
in_files = in_files[::sample]
# print len(in_files)
# optical_flow = cv2.DualTVL1OpticalFlow_create()
out_dir_u_curr = os.path.join(out_dir_u, video_name)
out_dir_v_curr = os.path.join(out_dir_v, video_name)
util.mkdir(out_dir_u_curr)
util.mkdir(out_dir_v_curr)
out_file_input = os.path.join(out_dir_u_curr, 'list_input.txt')
util.writeFile(out_file_input, in_files)
out_file_final_u = os.path.join(out_dir_u, video_name,
os.path.split(in_files[-2])[1])
out_file_final_v = os.path.join(out_dir_v, video_name,
os.path.split(in_files[-2])[1])
# print out_file_final_u
if os.path.exists(out_file_final_u) and os.path.exists(out_file_final_v):
print video_name, 'DONE already'
return 1
# return 0
command = [
'./optical_flow_try_2', out_file_input, out_dir_u_curr, out_dir_v_curr
]
command = ' '.join(command)
# print command
# t = time.time()
os.popen(command)
def test_resize_kp():
dir_meta = '../data/emotionet'
im_size = [256, 256]
out_dir_im = os.path.join(
dir_meta, 'preprocess_im_' + str(im_size[0]) + '_color_nodetect')
out_dir_kp = out_dir_im.replace('_im_', '_kp_')
out_dir_im_org = os.path.join(dir_meta, 'im')
im_file_list = out_dir_im + '_list_1.txt'
all_im = util.readLinesFromFile(im_file_list)
out_dir_scratch = '../scratch/emotionet_kp'
util.mkdir(out_dir_scratch)
for idx_im_curr, im_curr in enumerate(all_im[:100]):
im_org_file = im_curr.replace(out_dir_im, out_dir_im_org)
kp_in_file = im_curr.replace(out_dir_im,
out_dir_kp).replace('.jpg', '.npy')
if not os.path.exists(kp_in_file):
print 'CONTINUING', kp_in_file
continue
im_org = scipy.misc.imread(im_org_file)
im = scipy.misc.imread(im_curr)
kp = np.load(kp_in_file)
kp_org = kp / float(im_size[0])
kp_org[:, 0] = kp_org[:, 0] * im_org.shape[1]
kp_org[:, 1] = kp_org[:, 1] * im_org.shape[0]
out_file = os.path.join(out_dir_scratch, str(idx_im_curr) + '.jpg')
save_im_kp(im, kp, out_file)
out_file_org = os.path.join(out_dir_scratch,
str(idx_im_curr) + '_org.jpg')
save_im_kp(im_org, kp_org, out_file_org)
visualize.writeHTMLForFolder(out_dir_scratch)
def script_save_train_test_files():
meta_dir = '../data/charades'
train_test_dir = os.path.join(meta_dir, 'train_test_files')
util.mkdir(train_test_dir)
anno_files = [
os.path.join(meta_dir, 'annos_test.npz'),
os.path.join(meta_dir, 'annos_train.npz')
]
# anno_dir = os.path.join(meta_dir, 'vgg16_rgb_features_npy')
# out_files = [os.path.join(train_test_dir, 'vgg_16_rgb_npy_test.txt'),
# os.path.join(train_test_dir, 'vgg_16_rgb_npy_train.txt')]
# anno_dir = os.path.join(meta_dir, 'i3d_rgb')
# out_files = [os.path.join(train_test_dir, 'i3d_rgb_npy_test.txt'),
# os.path.join(train_test_dir, 'i3d_rgb_npy_train.txt')]
# anno_dir = os.path.join(meta_dir, 'i3d_flow')
# out_files = [os.path.join(train_test_dir, 'i3d_flow_npy_test.txt'),
# os.path.join(train_test_dir, 'i3d_flow_npy_train.txt')]
# anno_dir = os.path.join(meta_dir, 'i3d_both')
# out_files = [os.path.join(train_test_dir, 'i3d_both_test_wmiss.txt'),
# os.path.join(train_test_dir, 'i3d_both_train_wmiss.txt')]
anno_dir = os.path.join(meta_dir, 'i3d_charades_both')
out_files = [
os.path.join(train_test_dir, 'i3d_charades_both_test.txt'),
os.path.join(train_test_dir, 'i3d_charades_both_train_wmiss.txt')
]
post_pend = '.npy'
#
# out_file =
for anno_file, out_file in zip(anno_files, out_files):
make_train_test_files(anno_file, anno_dir, out_file, post_pend)
def script_download_image():
# idx_url_file,url_file,url_files,str_replace):
dir_meta = '../data/emotionet'
out_dir_im = os.path.join(dir_meta, 'im')
util.mkdir(out_dir_im)
str_replace = [
'http://cbcsnas01.ece.ohio-state.edu/EmotioNet/Images', out_dir_im
]
dir_url_files = os.path.join(dir_meta, 'emotioNet_challenge_files_server')
url_files = glob.glob(os.path.join(dir_url_files, '*.txt'))
url_files.sort()
args = []
out_files = []
for idx_url_file, url_file in enumerate(url_files):
print 'On file %d of %d' % (idx_url_file, len(url_files))
im = [
line_curr.split('\t')[0]
for line_curr in util.readLinesFromFile(url_file)
]
# out_files = [im_curr.replace(str_replace[0],str_replace[1]) for im_curr in im]
for idx_im_curr, im_curr in enumerate(im):
out_file_curr = im_curr.replace(str_replace[0], str_replace[1])
if os.path.exists(out_file_curr):
out_files.append(out_file_curr)
continue
out_dir_curr = os.path.split(out_file_curr)[0]
util.makedirs(out_dir_curr)
args.append((im_curr, out_file_curr, idx_im_curr))
print len(args)
print url_file
print len(args)
print len(out_files)
return out_files
def detect_frames_single_head():
dir_meta = '../data/mmi'
dir_server = '/disk3'
str_replace = ['..', os.path.join(dir_server, 'maheen_data/eccv_18')]
click_str = 'http://vision3.idav.ucdavis.edu:1000'
anno_file_new = os.path.join(dir_meta, 'annos_all_oriented.txt')
im_size = [110, 110]
savegray = True
out_dir_im = os.path.join(dir_meta, 'preprocess_im_gray_110')
util.mkdir(out_dir_im)
str_replace_dir = [os.path.join(dir_meta, 'Sessions'), out_dir_im]
annos_all = util.readLinesFromFile(anno_file_new)
file_pairs = []
for idx_anno_curr, anno_curr in enumerate(annos_all):
im_file, view, emo = anno_curr.split(' ')
out_file = im_file.replace(str_replace_dir[0], str_replace_dir[1])
if os.path.exists(out_file) and int(view) != 2:
continue
out_dir_curr = os.path.split(out_file)[0]
util.makedirs(out_dir_curr)
file_pairs.append((im_file, out_file))
args = []
chunk_size = 50
chunks = [
file_pairs[x:x + chunk_size]
for x in range(0, len(file_pairs), chunk_size)
]
args = [(chunk_curr, im_size, savegray, idx_im_in, True)
for idx_im_in, chunk_curr in enumerate(chunks)]
print len(args)
for arg in args:
print arg[0]
size = preprocess_bp4d.saveCroppedFace_NEW_batch(arg)
def rough_work():
# input = io.imread('../data/bp4d/BP4D/BP4D-training/F001/T1/2440.jpg')
dir_im = '../data/bp4d/preprocess_im_110_color/F001/T1'
im_list = glob.glob(os.path.join(dir_im, '*.jpg'))
# print im_list
im_list = im_list[:10]
preds = fa.get_landmarks_simple(im_list)
out_dir_check = os.path.join('../scratch/check_kp_fan')
util.mkdir(out_dir_check)
for idx_im_curr, im_curr in enumerate(im_list):
im_curr = scipy.misc.imread(im_curr)
pts_rel = preds[idx_im_curr]
for pt_curr in pts_rel:
cv2.circle(im_curr, (int(pt_curr[0]), int(pt_curr[1])), 2,
(255, 255, 255), -1)
out_file_curr = os.path.join(out_dir_check, str(idx_im_curr) + '.jpg')
scipy.misc.imsave(out_file_curr, im_curr)
visualize.writeHTMLForFolder(out_dir_check)
def make_folds():
dir_meta = '../data/disfa'
dir_im_meta = os.path.join(dir_meta, 'Videos_LeftCamera_frames_200')
dir_anno_meta = os.path.join(dir_meta, 'ActionUnit_Labels')
# num_folds = 10
# out_dir_folds = os.path.join(dir_meta,'folds_'+str(num_folds))
num_folds = 3
out_dir_folds = os.path.join(dir_meta, 'folds_' + str(num_folds))
util.mkdir(out_dir_folds)
list_subs = glob.glob(os.path.join(dir_anno_meta, '*'))
list_subs = [os.path.split(sub)[1] for sub in list_subs]
list_subs.sort()
all_folds = []
for idx_fold in range(num_folds):
all_folds.append(list_subs[idx_fold::num_folds])
for idx_fold in range(num_folds):
train_file = os.path.join(out_dir_folds,
'train_' + str(idx_fold) + '.txt')
test_file = os.path.join(out_dir_folds,
'test_' + str(idx_fold) + '.txt')
train_fold = []
for i in range(len(all_folds)):
if i != idx_fold:
train_fold = train_fold + all_folds[i]
test_fold = all_folds[idx_fold]
print idx_fold, len(train_fold), len(test_fold)
assert len(
set(train_fold + test_fold)) == len(train_fold + test_fold) == 27
util.writeFile(train_file, train_fold)
util.writeFile(test_file, test_fold)
def make_train_test_subs():
dir_meta = '../data/bp4d'
im_dir_meta = os.path.join(dir_meta, 'BP4D', 'BP4D-training')
out_dir_subs = os.path.join(dir_meta, 'subs')
util.mkdir(out_dir_subs)
subs = [
os.path.split(dir_curr)[1]
for dir_curr in glob.glob(os.path.join(im_dir_meta, '*'))
]
print subs
print len(subs)
subs.sort()
print subs
num_splits = 3
folds = []
for fold_num in range(num_splits):
fold_curr = subs[fold_num::num_splits]
folds.append(fold_curr)
for fold_num in range(num_splits):
train_folds = []
for idx_fold, fold_curr in enumerate(folds):
if idx_fold != fold_num:
train_folds = train_folds + fold_curr
test_folds = folds[fold_num]
out_file_train = os.path.join(out_dir_subs,
'train_' + str(fold_num) + '.txt')
out_file_test = os.path.join(out_dir_subs,
'test_' + str(fold_num) + '.txt')
assert len(train_folds) + len(test_folds) == len(
list(set(train_folds + test_folds)))
print fold_num, len(train_folds), len(test_folds)
print out_file_train, out_file_test
util.writeFile(out_file_train, train_folds)
util.writeFile(out_file_test, test_folds)
def graph_overfitting():
log_file_us = '../experiments/graph_multi_video_with_L1_retF/graph_multi_video_with_L1_retF_aft_nonlin_RL_L2_non_lin_None_sparsify_0.5_graph_size_2_sigmoid_False_graph_sum_True_deno_8_n_classes_20_in_out_2048_1024_feat_dim_2048_1024_feat_ret_True_method_cos_ucf/all_classes_False_just_primary_False_limit_None_cw_True_MultiCrossEntropyMultiBranchWithL1_CASL_500_step_250_0.1_0.001_0.001_0.001_lw_1.00_1.00_0.50__noLimit/log_det.txt'
log_file_them = '../experiments/graph_multi_video_with_L1_retW/graph_multi_video_with_L1_retW_aft_nonlin_RL_L2_non_lin_None_sparsify_0.5_graph_size_2_sigmoid_False_graph_sum_True_deno_8_n_classes_20_in_out_2048_1024_feat_dim_2048_1024_feat_ret_True_method_cos_ucf/all_classes_False_just_primary_False_limit_None_cw_True_MultiCrossEntropyMultiBranchWithL1_CASL_250_step_250_0.1_0.001_0.001_0.001_lw_1.00_1.00_0.50__caslexp/log_det.txt'
log_file_us = '../experiments/graph_multi_video_with_L1_retF_tanh/graph_multi_video_with_L1_retF_tanh_aft_nonlin_RL_L2_non_lin_None_sparsify_percent_0.5_graph_size_1_sigmoid_True_graph_sum_True_deno_8_n_classes_20_in_out_2048_1024_feat_dim_2048_1024_feat_ret_True_method_cos_ucf/all_classes_False_just_primary_False_limit_None_cw_False_MultiCrossEntropyMultiBranchWithL1_CASL_250_step_250_0.1_0.001_0.001_0.001_lw_1.00_1.00_1.00_forplot_0/log_det.txt'
log_file_them = '../experiments/graph_multi_video_with_L1_retW_new/graph_multi_video_with_L1_retW_new_aft_nonlin_RL_L2_non_lin_None_sparsify_percent_0.5_graph_size_1_sigmoid_True_graph_sum_True_deno_8_n_classes_20_in_out_2048_1024_feat_dim_2048_1024_feat_ret_True_method_cos_ucf/all_classes_False_just_primary_False_limit_None_cw_False_MultiCrossEntropyMultiBranchWithL1_CASL_250_step_250_0.1_0.001_0.001_0.001_lw_1.00_1.00_1.00_actuallytanh_0/log_det.txt'
files = [log_file_us, log_file_them]
x_vals = range(0, 251, 10)
xAndYs = []
for file_curr in files:
lines = util.readLinesFromFile(file_curr)
lines = lines[:26]
print lines[0]
det_vals = []
# for line in lines:
# line = [val for val in line.split(' ') if val is not '']
# print line
# det_vals.append(float(line[-1]))
# raw_input()
det_vals = [float(line.split('\t')[-1]) for line in lines]
# det_vals = det_vals[::5]
xAndYs.append((x_vals, det_vals))
out_file = '../scratch/qualitative_figs_wacv/graph_overfitting.jpg'
util.mkdir('../scratch/qualitative_figs_wacv')
legend_entries = ['Ours-MCASL', 'CASL-Graph']
xlabel = 'Training Epoch'
ylabel = 'Detection Accuracy'
title = 'Detection Accuracy at 0.5 Overlap'
visualize.plotSimple(xAndYs,
out_file=out_file,
xlabel=xlabel,
ylabel=ylabel,
legend_entries=legend_entries,
title=title)
def main():
out_dir = '../scratch/check_first_frames_with_cam_on'
out_dir_txt = '../scratch/check_first_frames_with_cam_on_txt'
out_file_offsets = '../metadata/fixing_offsets_with_cam_on/video_offsets_manual.csv'
util.mkdir(out_dir_txt)
times_file = os.path.join(out_dir, 'times.npy')
im_files_to_check = util.readLinesFromFile(
os.path.join(out_dir, 'manual_check.txt'))
args = []
out_files = []
for im_file in im_files_to_check:
out_file = os.path.join(
out_dir_txt,
os.path.split(im_file)[1].replace('.jpg', '.txt'))
out_files.append(out_file)
if not os.path.exists(out_file):
args.append((im_file, out_file))
print len(args), len(im_files_to_check)
# record_offsets(args)
# fix_offset_files(out_files)
lines = ['im_file,offset']
for im_file, offset_file in zip(im_files_to_check, out_files):
offset = util.readLinesFromFile(offset_file)[0]
# sanity check
num = int(offset)
lines.append(im_file + ',' + offset)
print lines
print len(lines)
print out_file_offsets
util.writeFile(out_file_offsets, lines)
def main(video_file, to_run, fps, smooth, post_dir):
print video_file
print 'to_run', to_run
print 'sec', fps
# video_file = '../data/Surveillance/ch02_20161212115300.mp4'
out_dir = video_file[:video_file.rindex('.')] + '_result_files' + post_dir
data_dir = video_file[:video_file.rindex('.')] + '_frames' + post_dir
util.mkdir(data_dir)
util.mkdir(out_dir)
# fps = 5 #extract one frame every n seconds
size_output = [416, 416]
if to_run == 'all' or to_run == 'extract':
print 'EXTRACTING FRAMES'
subprocess.call('rm ' + os.path.join(data_dir, '*.jpg'), shell=True)
extract_frames(video_file, data_dir, fps, size_output)
visualize.writeHTMLForFolder(data_dir)
print 'DONE EXTRACTING FRAMES'
# t = time.time()
if to_run == 'all' or to_run == 'test':
print 'TESTING FRAMES'
test_frames(out_dir, data_dir)
print 'DONE TESTING'
if to_run == 'all' or to_run == 'graph':
print 'PLOTTING DETECTIONS OVER TIME'
plot_detections_over_time(data_dir, out_dir, fps, smooth)
# print time.time()-t
if to_run == 'plot':
print data_dir
plot_detections(data_dir, out_dir)
def main():
fps = 10
width = 672
height = 380
str_dir = '_'.join([str(val) for val in [width,height,fps]])
out_dir_testing = '../data/pain_no_pain_x2h_intervals_for_extraction_'+str_dir+'fps'
out_dir_flow = '../data_other/pain_no_pain_x2h_intervals_for_extraction_'+str_dir+'fps'
util.mkdir(out_dir_testing)
ofe = OpticalFlowExtractor(output_dir = out_dir_testing, num_processes = multiprocessing.cpu_count(), output_flow = out_dir_flow)
# # extract optical flow
# # ofe.extract_frames(replace = False, subjects_to_extract = None)
# # add symlinks if storing optical flow in a different drive
# ofe.add_symlinks(subjects_to_extract = None)
# # collate magnitudes and files names in text files. need to manually run commands files after this step.
# ofe.collate_magnitude(subjects_to_extract = ['inkasso'])
# ['julia', 'kastanjett', 'naughty_but_nice', 'sir_holger'])
# create csv with thresholded images only
# ofe.create_thresholded_csv(thresh = 0.7,subjects_to_extract = ['inkasso'])
ofe.create_thresholded_csv(thresh = 0.01,subjects_to_extract = None, percent = True)
def script_viz_k_means():
out_dir_htmls = '../experiments/figures/primary_caps_viz_pca'.replace(
str_replace[0], str_replace[1])
util.mkdir(out_dir_htmls)
out_dir_im = os.path.join(out_dir_htmls, 'im')
util.mkdir(out_dir_im)
caps, test_file, convnet, imsize = get_caps_compiled()
num_clusters = 32
# arr_vals = [(x,y,filter_num) for x in range(6) for y in range(6) for filter_num in range(32)]
arr_vals = [(x, y, filter_num) for x in [3] for y in [5]
for filter_num in [3]]
test_im = [
scipy.misc.imread(line_curr.split(' ')[0])
for line_curr in util.readLinesFromFile(test_file)
]
print len(test_im)
print test_im[0].shape
for x, y, filter_num in arr_vals:
out_dir_curr = os.path.join(
out_dir_im,
str(x) + '_' + str(y) + '_' + str(filter_num))
util.mkdir(out_dir_curr)
out_file_html = os.path.join(
out_dir_htmls,
str(x) + '_' + str(y) + '_' + str(filter_num) + '.html')
# if os.path.exists(out_file_html):
# continue
pca(caps,
num_clusters,
filter_num,
x,
y,
test_im,
out_dir_curr,
out_file_html,
convnet,
imsize,
rewrite=False)
# break
visualize.writeHTMLForFolder(out_dir_im)
def script_viz_mag():
out_dir_htmls = '../experiments/figures/primary_caps_viz'.replace(
str_replace[0], str_replace[1])
util.mkdir(out_dir_htmls)
out_dir_im = os.path.join(out_dir_htmls, 'im')
util.mkdir(out_dir_im)
caps, test_file, convnet, imsize = get_caps_compiled()
mags = np.linalg.norm(caps, axis=4)
print mags.shape
print np.min(mags), np.max(mags)
test_im = [
scipy.misc.imread(line_curr.split(' ')[0])
for line_curr in util.readLinesFromFile(test_file)
]
print len(test_im)
print test_im[0].shape
for x in range(mags.shape[2]):
for y in range(mags.shape[3]):
out_file_html = os.path.join(out_dir_htmls,
str(x) + '_' + str(y) + '.html')
ims_html = []
captions_html = []
for filter_num in range(mags.shape[1]):
out_dir_curr = os.path.join(
out_dir_im,
str(x) + '_' + str(y) + '_' + str(filter_num))
util.mkdir(out_dir_curr)
im_row, caption_row = save_ims(mags, filter_num, x, y, test_im,
out_dir_curr, convnet, imsize)
im_row = [
util.getRelPath(im_curr, dir_server) for im_curr in im_row
]
# caption_row = [os.path.split(im_curr)[1][:-4] for im_curr in im_row]
ims_html.append(im_row[:10] + im_row[-10:])
captions_html.append(caption_row[:10] + caption_row[-10:])
visualize.writeHTML(out_file_html, ims_html, captions_html, 40, 40)
def save_npys():
out_dir_features = os.path.join(dir_meta, 'i3d')
util.mkdir(out_dir_features)
out_dir_train = os.path.join(out_dir_features, 'train_data')
out_dir_val = os.path.join(out_dir_features, 'val_data')
util.mkdir(out_dir_train)
util.mkdir(out_dir_val)
out_dirs = [out_dir_train, out_dir_val]
features = get_i3d()
ids, durations, train_val, labels_all, times_all, labels = get_rel_anno_info(
)
assert len(features) == len(ids)
train_val = np.array(train_val)
train_val_bool = np.zeros(train_val.shape).astype(int)
train_val_bool[train_val == 'validation'] = 1
for idx_id_curr, id_curr in enumerate(ids):
out_file_curr = os.path.join(out_dirs[train_val_bool[idx_id_curr]],
id_curr + '.npy')
if os.path.exists(out_file_curr):
continue
features_curr = features[idx_id_curr]
duration_curr = durations[idx_id_curr]
feature_len = features_curr.shape[0]
pred_len = duration_curr * 25 // 16
diff = np.abs(pred_len - feature_len)
# diffs.append(diff)
if diff > 2:
# print 'Continuing',diff, feature_len, pred_len, duration_curr, train_val_bool[idx_id_curr]
print id_curr, train_val_bool[idx_id_curr]
continue
def main():
# out_dir_meta = '../experiments/bl_khorrami_ck_96_nobn_pixel_augment_255_range'
# range_splits = [0,1,2,3,4,5]
out_dir_meta = '../experiments/bl_khorrami_ck_96_nobn_pixel_augment_255_range_trans_fix'
range_splits = range(6, 10)
print range_splits
# range(10)
util.mkdir(out_dir_meta)
all_accuracy = []
for split_num in range_splits:
train_file = '../data/ck_96/train_test_files/train_' + str(
split_num) + '.txt'
test_file = '../data/ck_96/train_test_files/test_' + str(
split_num) + '.txt'
mean_file = '../data/ck_96/train_test_files/train_' + str(
split_num) + '_mean.png'
std_file = '../data/ck_96/train_test_files/train_' + str(
split_num) + '_std.png'
list_of_to_dos = ['flip', 'rotate', 'scale_translate', 'pixel_augment']
mean_im = scipy.misc.imread(mean_file).astype(np.float32)
std_im = scipy.misc.imread(std_file).astype(np.float32)
batch_size = 128
batch_size_val = None
num_epochs = 500
save_after = 100
disp_after = 1
plot_after = 10
test_after = 1
lr = [0.001, 0.001]
# lr = [0.0001,0.0001]
dec_after = 300
model_name = 'khorrami_ck_96'
criterion = nn.CrossEntropyLoss()
gpu_id = 0
num_workers = 2
model_num = num_epochs - 1
# model_file = None
# epoch_start = 0
# lr_dir_train = lr
lr_dir_train = [0.01, 0.01]
# strs_append = '_'.join([str(val) for val in [num_epochs,dec_after,lr_dir_train[0],lr_dir_train[1],'100_dec']])
strs_append = '_'.join([
str(val) for val in
[num_epochs, dec_after, lr_dir_train[0], lr_dir_train[1]]
])
out_dir_train = os.path.join(
out_dir_meta, 'split_' + str(split_num) + '_' + strs_append)
print out_dir_train
epoch_start = 401
strs_append = '_'.join(
[str(val) for val in [400, 300, lr_dir_train[0], lr_dir_train[1]]])
out_dir_res = os.path.join(
out_dir_meta, 'split_' + str(split_num) + '_' + strs_append)
# strs_append = '_'.join([str(val) for val in [250,200,lr_dir_train[0],lr_dir_train[1]]])
# model_file = os.path.join(out_dir_meta,'split_'+str(split_num)+'_'+strs_append,'model_200.pt')
model_file = os.path.join(out_dir_res, 'model_399.pt')
# raw_input()
util.mkdir(out_dir_train)
data_transforms = {}
data_transforms['train'] = transforms.Compose([
lambda x: augment_image(
x, list_of_to_dos, mean_im=mean_im, std_im=std_im),
transforms.ToTensor(), lambda x: x * 255.
])
data_transforms['val'] = transforms.Compose(
[transforms.ToTensor(), lambda x: x * 255.])
train_data = dataset.CK_96_Dataset(train_file, mean_file, std_file,
data_transforms['train'])
test_data = dataset.CK_96_Dataset(test_file, mean_file, std_file,
data_transforms['val'])
train_model(out_dir_train,
train_data,
test_data,
batch_size=batch_size,
batch_size_val=batch_size_val,
num_epochs=num_epochs,
save_after=save_after,
disp_after=disp_after,
plot_after=plot_after,
test_after=test_after,
lr=lr,
dec_after=dec_after,
model_name=model_name,
criterion=criterion,
gpu_id=gpu_id,
num_workers=num_workers,
model_file=model_file,
epoch_start=epoch_start)
test_model(out_dir_train,
model_num,
train_data,
test_data,
model_name=model_name,
batch_size_val=batch_size_val,
criterion=criterion)
res_dir = os.path.join(out_dir_train,
'results_model_' + str(model_num))
log_file = os.path.join(res_dir, 'log.txt')
accuracy = util.readLinesFromFile(log_file)[-1]
accuracy = float(accuracy.split(' ')[1])
all_accuracy.append(accuracy)
print all_accuracy, np.mean(all_accuracy), np.std(all_accuracy)
def save_class_vary_attr(out_dir_train,
model_num,
train_data,
test_data,
gpu_id=0,
model_name='alexnet',
batch_size_val=None,
criterion=nn.CrossEntropyLoss(),
margin_params=None,
network_params=None,
barebones=True,
class_rel=0,
au=False):
mag_range = np.arange(-0.5, 0.6, 0.1)
out_dir_results = os.path.join(
out_dir_train, 'save_class_vary_attr_single_batch_' + str(model_num) +
'_' + str(class_rel))
util.makedirs(out_dir_results)
out_dir_im = os.path.join(out_dir_results, 'im_save_' + str(class_rel))
util.mkdir(out_dir_im)
print out_dir_results
model_file = os.path.join(out_dir_train, 'model_' + str(model_num) + '.pt')
log_arr = []
# network = models.get(model_name,network_params)
if batch_size_val is None:
batch_size_val = len(test_data)
test_dataloader = torch.utils.data.DataLoader(test_data,
batch_size=batch_size_val,
shuffle=False,
num_workers=1)
torch.cuda.device(0)
iter_begin = 0
model = torch.load(model_file)
model.cuda()
model.eval()
predictions = []
labels_all = []
out_all = []
caps_all = []
recons_all = {}
mean_im = test_data.mean
mean_im = mean_im[np.newaxis, :, :]
std_im = test_data.std[np.newaxis, :, :]
ims_all = []
for num_iter, batch in enumerate(test_dataloader):
print 'NUM_ITER', num_iter
labels = batch['label'].cpu().numpy()
data = batch['image'].cpu().numpy()
if au:
print labels.shape
rel_vals = labels[:, class_rel] > 0
print np.sum(rel_vals)
labels = np.zeros((np.sum(rel_vals), labels.shape[1]))
labels[:, class_rel] = 1
print labels.shape
print rel_vals.shape
else:
rel_vals = labels == class_rel
labels = labels[rel_vals]
data = data[rel_vals]
batch['image'] = torch.Tensor(data)
batch['label'] = torch.LongTensor(labels)
print labels.shape
print data.shape
if data.shape[0] == 1:
continue
# raw_input()
# batch = test_dataloader.next()
if criterion == 'marginmulti':
labels = Variable(batch['label'].float().cuda())
else:
labels = Variable(torch.LongTensor(batch['label']).cuda())
# labels_all.append(batch['label'].numpy())
data = Variable(batch['image'].cuda())
# recons_all[(0,0)] = data.data.cpu().numpy()
# labels = Variable(torch.LongTensor(batch['label']).cuda())
# x = model.features(data)
# _,routes = model.caps.forward_intrusive(x)
# routes = [np.squeeze(routes_curr) for routes_curr in routes]
# print len(routes), routes[0].shape
# output, caps =
classes, reconstructions_gt, _, caps = model(data,
labels,
return_caps=True)
caps_mag = (caps**2).sum(dim=-1)**0.5
caps_unit = caps / caps_mag.view(caps_mag.size(0), caps_mag.size(1), 1)
# recons_all_all = []
recons_all = []
for attr_num in range(32):
for mag_curr in mag_range:
caps_curr = caps_unit.clone()
caps_curr[:, :, attr_num] = mag_curr
caps_mag_curr = (caps_curr**2).sum(dim=-1)**0.5
caps_curr = caps_curr / caps_mag_curr.view(
caps_mag_curr.size(0), caps_mag_curr.size(1), 1)
caps_curr = caps_curr * caps_mag.view(caps_mag.size(0),
caps_mag.size(1), 1)
recon_curr = model.just_reconstruct(caps_curr, labels)
recons_all.append(recon_curr)
# recons_all_all.append(recons_all)
# print caps.size()
# raw_input()
classes, reconstructions_gt, caps = [
val.data.cpu().numpy()
for val in [classes, reconstructions_gt, caps]
]
recons_all = [val.data.cpu().numpy() for val in recons_all]
labels = labels.data.cpu().numpy()
preds = np.argmax(classes, 1)
# print preds.shape, labels.shape
# print np.sum(preds==labels)/float(labels.size)
batch_size = data.shape[0]
data = data.data.cpu().numpy()
# ims_all = []
# for label_curr in range(8):
# rel_idx = np.where(labels==label_curr)
# print rel_idx
# ims_all_all = [[] for i in range(8)]
# ims_all = []
for im_num in range(batch_size):
for attr_num in range(32):
post_pend = []
im_out = []
gt_label = labels[im_num]
pred_label = preds[im_num]
# ims_all = ims_all_all[gt_label]
# print 'gt',gt_label,'pred',pred_label
im_in = data[im_num]
# [0]
# print im_in.shape
# raw_input()
im_in = (im_in * std_im) + mean_im
# print im_in.shape
im_out.append(im_in)
post_pend.append(['org'])
# recon_gt = reconstructions_gt[im_num]
# # [0]
# recon_gt = (recon_gt*std_im)+mean_im
# # print recon_gt.shape
# im_out.append(recon_gt)
# post_pend.append(['recon_gt'])
# routes_im = [np.sum(route[:,im_num,:,:],2) for route in routes]
# # for val in im_out:
# # print val.shape
# for label_curr in range(len(recons_all)):
im_out.append(im_in)
post_pend.append(['org'])
for idx_mag_curr, mag_curr in enumerate(mag_range):
idx_curr = attr_num * len(mag_range) + idx_mag_curr
# print idx_curr,len(recons_all)
recon_rel = np.array(recons_all[idx_curr][im_num])
recon_rel = (recon_rel * std_im) + mean_im
# recon_rel = recon_rel+np.min(recon_rel)
# recon_rel = recon_rel/np.max(recon_rel)
im_out.append(recon_rel)
post_pend.append([attr_num, idx_mag_curr])
# pre_vals = [im_num,gt_label,pred_label]
pre_vals = [num_iter, im_num]
ims_row = save_all_im(out_dir_im, pre_vals, im_out, post_pend)
ims_all.append(ims_row)
# print ims_all
# print len(ims_all)
# print len(ims_all[0])
# raw_input()
break
# mats_to_save = []
# mats_to_save = [labels,preds,routes[0],routes[1]]
# mats_names = ['labels','preds','routes_0','routes_1']
# for mat_curr, file_curr in zip(mats_to_save,mats_names):
# out_file_curr = os.path.join(out_dir_results,file_curr+'.npy')
# np.save(out_file_curr,mat_curr)
np.save(os.path.join(out_dir_results, 'ims_all.npy'), np.array(ims_all))
