def load_train_val_files():
datapath ='/research2/ECCV_dataset_resized'
val_data = [11,16,21,22,33,36,38,53,59,92]
dataset={}
train_input=[]
train_gt=[]
val_input=[]
val_gt=[]
folders = glob.glob(os.path.join(datapath,'*'))
folders = natsorted(folders)
count =1
for ff in range(len(folders)):
subfolders = glob.glob(os.path.join(datapath,folders[ff],'*'))
subfolders = natsorted(subfolders)
for ss in range(len(subfolders)):
print('folder:%d subfolder %d \n' %(ff,ss))
files = glob.glob(os.path.join(datapath,folders[ff],subfolders[ss],'*.bmp'))
files = natsorted(files)
if count not in val_data:
train_input.append(files[:-1])
train_gt.append(files[-1])
else:
val_input.append(files[:-1])
val_gt.append(files[-1])
count += 1
dataset['train_input'] = train_input
dataset['train_gt'] = train_gt
dataset['val_input'] = val_input
dataset['val_gt']= val_gt
with open('ECCV_256size.pickle','wb') as f:
pickle.dump(dataset,f)
def draw_mean(csvpath):
min_idx = 16
files = glob.glob(os.path.join(csvpath, '*.csv'))
files = natsorted(files)
num_files = len(files)
color = iter(cm.rainbow(np.linspace(0, 1, num_files)))
mean_error = []
mean_ang = []
mean_deg = []
for f in range(num_files):
err, ang, deg = mean_read_files(files[f], csvpath)
mean_error.append(err)
mean_ang.append(ang)
mean_deg.append(deg)
if len(err) < min_idx - 1:
print('Minimum epochs is %d, current file has %d epochs' %
(min_idx, len(err)))
pdb.set_trace()
for d in range(num_files):
labelname = files[d]
labelname = labelname.split('/')
labelname = labelname[-1].split('.')
c = next(color)
plt.plot(range(min_idx),
mean_error[d][:min_idx],
color=c,
label=labelname[0])
plt.legend(loc='best')
#plt.legend(loc='upper left',bbox_to_anchor=[0,1],ncol=1)
plt.savefig(os.path.join(csvpath, 'mean_err.png'))
plt.close()
color = iter(cm.rainbow(np.linspace(0, 1, num_files)))
for d in range(num_files):
labelname = files[d]
labelname = labelname.split('/')
labelname = labelname[-1].split('.')
c = next(color)
plt.plot(range(min_idx),
mean_ang[d][:min_idx],
color=c,
label=labelname[0])
plt.legend(loc='upper left', bbox_to_anchor=[0, 1], ncol=1)
plt.savefig(os.path.join(csvpath, 'mean_ang.png'))
plt.close()
color = iter(cm.rainbow(np.linspace(0, 1, num_files)))
for d in range(num_files):
labelname = files[d]
labelname = labelname.split('/')
labelname = labelname[-1].split('.')
c = next(color)
plt.plot(range(min_idx),
mean_deg[d][:min_idx],
color=c,
label=labelname[0])
plt.legend(loc='upper left', bbox_to_anchor=[0, 1], ncol=1)
plt.savefig(os.path.join(csvpath, 'good_deg.png'))
plt.close()
def diff_lights(mainpath):
sum_diff = 0.0
num = 0
files = glob.glob(os.path.join(mainpath, '*.csv'))
files = natsorted(files)
for d in range(len(files)):
print('dir %d /%d \n' % (d, len(files)))
file_ = os.path.join(mainpath, files[d])
with open(file_, 'r') as f:
content = f.readlines()
for line in range(1, len(content)):
values = content[line].strip().split(",")
sum_diff += float(values[4])
num += 1
print('mean different lights: %0.6f \n' % (sum_diff / num))
def finding_bestview(mainpath):
min_err = 100.0
min_arg = 100.0
max_deg = 0.0
err_view = 'err'
arg_view = 'arg'
deg_view = 'deg'
files = glob.glob(os.path.join(mainpath, '*.csv'))
files = natsorted(files)
for d in range(len(files)):
print('dir %d /%d \n' % (d, len(files)))
file_ = os.path.join(mainpath, files[d])
#[min_err,min_arg,max_deg,arr_epoch,arg_epoch,deg_epoch,err_view,arg_view,deg_view] = searchminmax(file_,min_err,min_arg,max_deg,files[f])
min_err, min_arg, max_deg, err_view, arg_view, deg_view = searchminmax(
file_, min_err, min_arg, max_deg, err_view, arg_view, deg_view,
file_)
print('min_err:%f view:%s \n' % (min_err, err_view))
print('min_arg:%f view:%s\n' % (min_arg, arg_view))
print('max_deg:%f view:%s\n' % (max_deg, deg_view))
def compute_average(path_, num_imgs):
main_path = os.path.join(
'/home/yjyoon/Dropbox/ECCV_result/avg_view/%03d/' % num_imgs)
if not os.path.exists(main_path):
os.makedirs(main_path)
dirs = os.listdir(path_)
for d in range(len(dirs)):
subp = os.path.join(path_, dirs[d])
subdirs = os.listdir(subp)
for subd in range(len(subdirs)):
images = glob.glob(
os.path.join(path_, dirs[d], subdirs[subd], '*.bmp'))
images = natsorted(images)
rand_idx = np.random.permutation(len(images))
avg_img = np.zeros((600, 800, 3)).astype(np.float)
for idx in range(num_imgs):
img = scipy.misc.imread(images[rand_idx[idx]])
avg_img += img
avg_img /= num_imgs
savepath = main_path + 'avg_img_%s_%03d.bmp' % (dirs[d],
int(subdirs[subd]))
scipy.misc.imsave(savepath, avg_img)
def drawplot(csvpath):
files = glob.glob(os.path.join(csvpath, '*.csv'))
files = natsorted(files)
num_files = len(files)
for f in range(num_files):
read_files(files[f], csvpath)
def main(_):
width_size = 905
height_size = 565
#width_size = 1104
#height_size = 764
#width_size = 1123
#height_size = 900
pp.pprint(flags.FLAGS.__flags)
if not os.path.exists(FLAGS.checkpoint_dir):
os.makedirs(FLAGS.checkpoint_dir)
if not os.path.exists(FLAGS.sample_dir):
os.makedirs(FLAGS.sample_dir)
if not os.path.exists('./logs_multifreq_noskip'):
os.makedirs('./logs_multifreq_noskip')
gpu_config = tf.GPUOptions(per_process_gpu_memory_fraction=FLAGS.gpu)
#with tf.Session() as sess:
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_config)) as sess:
if FLAGS.is_train:
dcgan = DCGAN(sess, image_size=FLAGS.image_size, batch_size=FLAGS.batch_size,\
num_block = FLAGS.num_block,dataset_name=FLAGS.dataset,is_crop=FLAGS.is_crop, checkpoint_dir=FLAGS.checkpoint_dir)
else:
dcgan = EVAL(sess, batch_size=1,num_block=FLAGS.num_block,ir_image_shape=[None,None,1],dataset_name=FLAGS.dataset,\
is_crop=False, checkpoint_dir=FLAGS.checkpoint_dir)
print('deep model test \n')
if FLAGS.is_train:
dcgan.train(FLAGS)
else:
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
print '1: Estimating Normal maps from arbitary obejcts \n'
print '2: EStimating Normal maps according to only object tilt angles(Light direction is fixed(EX:3) \n'
print '3: Estimating Normal maps according to Light directions and object tilt angles \n'
x = input('Selecting a Evaluation mode:')
VAL_OPTION = int(x)
if VAL_OPTION == 1: # arbitary dataset
print("Computing arbitary dataset ")
trained_models = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
trained_models = natsorted(trained_models)
model = trained_models[6]
model = model.split('/')
model = model[-1]
print('Load trained network: %s\n' % model)
dcgan.load(FLAGS.checkpoint_dir, model)
datapath = '/research3/datain/gmchoe_normal/0403/IR_0.25'
#datapath = '/research3/dataout/ECCV_2018/2130/centerview'
#datapath = '/research2/proposal_linux/dataset/coin'
savepath = datapath
mean_nir = -0.3313
img_files = glob.glob(os.path.join(datapath, '*.png'))
img_files = natsorted(img_files)
pdb.set_trace()
#listdir = natsorted(os.listdir(datapath))
#fulldatapath = natsorted(fulldatapath)
for idx in xrange(0, len(img_files)):
print('Processing %d/%d \n' % (len(img_files), idx))
#img_file = glob.glob(os.path.join(datapath,'nir.jpg'))
#img_file = glob.glob(os.path.join(datapath,listdir[idx]))
input_ = scipy.misc.imread(img_files[idx],
'F').astype(float)
height_size = input_.shape[0]
width_size = input_.shape[1]
#input_ = scipy.misc.imresize(input_,[565,905])
input_ = np.reshape(
input_,
(height_size, width_size, 1)) # LF size:383 x 552
#input_ = np.power(input_,0.6)
nondetail_input_ = ndimage.gaussian_filter(input_,
sigma=(1, 1, 0),
order=0)
input_ = input_ / 127.5 - 1.0
nondetail_input_ = nondetail_input_ / 127.5 - 1.0 # normalize -1 ~1
detail_input_ = input_ - nondetail_input_
nondetail_input_ = np.reshape(
nondetail_input_,
(1, height_size, width_size, 1)) # LF size:383 x 552
detail_input_ = np.reshape(detail_input_,
(1, height_size, width_size, 1))
#detail_input_ = detail_input_/127.5 -1.0 # normalize -1 ~1
start_time = time.time()
sample = sess.run(dcgan.G,
feed_dict={
dcgan.nondetail_images:
nondetail_input_,
dcgan.detail_images: detail_input_
})
print('time: %.8f' % (time.time() - start_time))
sample = np.squeeze(sample).astype(np.float32)
# normalization #
output = np.sqrt(np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output = (output + 1.) / 2.
"""
if not os.path.exists(os.path.join(savepath,'%s/%s/%s' %(FLAGS.dataset,model,listdir[idx]))):
os.makedirs(os.path.join(savepath,'%s/%s/%s' %(FLAGS.dataset,model,listdir[idx])))
"""
savename = os.path.join(
savepath, 'result/%s.bmp' % (img_files[idx][-10:]))
#savename = os.path.join(savepath,'single_normal_%02d.bmp' % (idx+1))
#savename = os.path.join(savepath,'%s/%s/%s/single_normal.bmp' % (FLAGS.dataset,model,listdir[idx]))
scipy.misc.imsave(savename, output)
elif VAL_OPTION == 2: # light source fixed
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
#save_files = glob.glob(os.path.join(FLAGS.checkpoint_dir,FLAGS.dataset,'DCGAN.model*'))
#save_files = natsorted(save_files)
savepath = './skip/L1_ang'
load = dcgan.load(FLAGS.checkpoint_dir)
if load:
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10):
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(img +
'/3.bmp').astype(float)
#gt_ = scipy.misc.imread('/research2/IR_normal_small/save016/1/12_Normal.bmp').astype(float)
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = np.reshape(input_, (600, 800, 1))
nondetail_input_ = ndimage.gaussian_filter(
input_, sigma=(1, 1, 0), order=0)
input_ = input_ / 127.5 - 1.0
nondetail_input_ = nondetail_input_ / 127.5 - 1.0 # normalize -1 ~1
detail_input_ = input_ - nondetail_input_
nondetail_input_ = np.reshape(
nondetail_input_, (1, 600, 800, 1))
detail_input_ = np.reshape(detail_input_,
(1, 600, 800, 1))
start_time = time.time()
sample = sess.run(
[dcgan.G],
feed_dict={
dcgan.nondetail_images: nondetail_input_,
dcgan.detail_images: detail_input_
})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.zeros((600, 800, 3)).astype(np.float32)
output[:, :, 0] = sample[:, :, 0] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 1] = sample[:, :, 1] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 2] = sample[:, :, 2] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[output == inf] = 0.0
sample = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%03d/%d' %
(list_val[idx], idx2))):
os.makedirs(
os.path.join(
savepath,
'%03d/%d' % (list_val[idx], idx2)))
savename = os.path.join(
savepath, '%03d/%d/single_normal.bmp' %
(list_val[idx], idx2))
scipy.misc.imsave(savename, sample)
else:
print("Failed to load network")
elif VAL_OPTION == 3: # depends on light sources
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
mean_nir = -0.3313 #-1~1
savepath = './angle_light_result'
if not os.path.exists(os.path.join(savepath)):
os.makedirs(os.path.join(savepath))
load = dcgan.load(FLAGS.checkpoint_dir)
if load:
print(" Load Success")
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10): #tilt angles 1~9
for idx3 in range(1, 13): # light source
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(
img + '/%d.bmp' % idx3).astype(
np.float32) #input NIR image
input_ = scipy.misc.imresize(
input_, [600, 800], 'nearest')
input_ = np.reshape(input_, (600, 800, 1))
nondetail_input_ = ndimage.gaussian_filter(
input_, sigma=(1, 1, 0), order=0)
input_ = input_ / 127.5 - 1.0
nondetail_input_ = nondetail_input_ / 127.5 - 1.0 # normalize -1 ~1
detail_input_ = input_ - nondetail_input_
nondetail_input_ = np.reshape(
nondetail_input_, (1, 600, 800, 1))
detail_input_ = np.reshape(
detail_input_, (1, 600, 800, 1))
start_time = time.time()
sample = sess.run(
[dcgan.G],
feed_dict={
dcgan.nondetail_images:
nondetail_input_,
dcgan.detail_images: detail_input_
})
sample = np.squeeze(sample[-1]).astype(
np.float32)
print('time: %.8f' %
(time.time() - start_time))
# normalization #
output = np.sqrt(
np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%s/%03d/%d' %
(FLAGS.dataset, list_val[idx],
idx2))):
os.makedirs(
os.path.join(
savepath, '%s/%03d/%d' %
(FLAGS.dataset, list_val[idx],
idx2)))
savename = os.path.join(
savepath,
'%s/%03d/%d/single_normal_%03d.bmp' %
(FLAGS.dataset, list_val[idx], idx2, idx3))
scipy.misc.imsave(savename, output)
else:
print("Failed to laod network")
def main(_):
pp.pprint(flags.FLAGS.__flags)
if not os.path.exists(FLAGS.checkpoint_dir):
os.makedirs(FLAGS.checkpoint_dir)
if not os.path.exists(FLAGS.sample_dir):
os.makedirs(FLAGS.sample_dir)
if not os.path.exists(os.path.join('./logs', time.strftime('%d%m'))):
os.makedirs(os.path.join('./logs', time.strftime('%d%m')))
gpu_config = tf.GPUOptions(per_process_gpu_memory_fraction=FLAGS.gpu)
#with tf.Session() as sess:
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_config)) as sess:
if FLAGS.is_train:
fcn = FCN(sess, image_size=FLAGS.image_size, batch_size=FLAGS.batch_size,\
dataset_name=FLAGS.dataset,is_crop=FLAGS.is_crop, checkpoint_dir=FLAGS.checkpoint_dir)
else:
fcn = EVAL(sess, batch_size=1,rgb_image_shape=[224,224,3],dataset_name=FLAGS.dataset,\
is_crop=False, checkpoint_dir=FLAGS.checkpoint_dir)
if FLAGS.is_train:
fcn.train(FLAGS)
else:
VGG_mean = [103.939, 116.779, 123.68]
train_rgb_data = open("db/Oxford_data1_RGB_train.txt")
train_rgblist = train_rgb_data.readlines()
train_depth_data = open("db/Oxford_data1_depth_train.txt")
train_depthlist = train_depth_data.readlines()
shuf = range(0, len(train_rgblist))
random.shuffle(shuf)
shuf = shuf[:10]
save_files = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'FCN.model*'))
save_files = natsorted(save_files)
savepath = './Depth_seg'
if not os.path.exists(os.path.join(savepath)):
os.makedirs(os.path.join(savepath))
model = save_files[-2]
model = model.split('/')
model = model[-1]
fcn.load(FLAGS.checkpoint_dir, model)
for m in range(len(shuf)):
rgbpath = train_rgblist[shuf[m]]
rgb_img = scipy.misc.imread(rgbpath[:-1]).astype(np.float32)
rgb_img = center_crop(rgb_img, 224)
rgb_img = np.reshape(rgb_img, (1, 224, 224, 3))
depthpath = train_depthlist[shuf[m]]
depth_img = sio.loadmat(depthpath[:-1])
depth_img = depth_img['depth']
depth_img = np.reshape(
depth_img, [depth_img.shape[0], depth_img.shape[1], 1])
depth_img = center_crop(depth_img, 224)
start_time = time.time()
predict = sess.run(fcn.pred_seg,
feed_dict={
fcn.rgb_images: rgb_img,
fcn.keep_prob: 1.0
})
predict = np.squeeze(predict).astype(np.float32)
print('time: %.8f' % (time.time() - start_time))
if not os.path.exists(os.path.join(savepath, '%s' % (model))):
os.makedirs(os.path.join(savepath, '%s' % (model)))
savename = os.path.join(
savepath, '%s/predict_%03d.jpg' % (model, shuf[m]))
scipy.misc.imsave(savename, predict.astype(np.uint8))
savename = os.path.join(savepath,
'%s/gt_%03d.jpg' % (model, shuf[m]))
scipy.misc.imsave(savename,
np.squeeze(depth_img).astype(np.uint8))
def main(_):
width_size = 905
height_size = 565
#width_size = 1104
#height_size = 764
#width_size = 1123
#height_size = 900
pp.pprint(flags.FLAGS.__flags)
if not os.path.exists(FLAGS.checkpoint_dir):
os.makedirs(FLAGS.checkpoint_dir)
if not os.path.exists(FLAGS.sample_dir):
os.makedirs(FLAGS.sample_dir)
gpu_config = tf.GPUOptions(per_process_gpu_memory_fraction=FLAGS.gpu)
#with tf.Session() as sess:
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_config)) as sess:
if FLAGS.is_train:
dcgan = DCGAN(sess, image_size=FLAGS.image_size, batch_size=FLAGS.batch_size,\
num_block = FLAGS.num_block,dataset_name=FLAGS.dataset,is_crop=FLAGS.is_crop, checkpoint_dir=FLAGS.checkpoint_dir)
else:
dcgan = EVAL(sess, batch_size=1,num_block=FLAGS.num_block,ir_image_shape=[None,None,1],dataset_name=FLAGS.dataset,\
is_crop=False, checkpoint_dir=FLAGS.checkpoint_dir)
print('deep model test \n')
if FLAGS.is_train:
dcgan.train(FLAGS)
else:
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
print '1: Estimating Normal maps from arbitary obejcts \n'
print '2: EStimating Normal maps of NIR dataset \n'
x = input('Selecting a Evaluation mode:')
VAL_OPTION = int(x)
if VAL_OPTION == 1: # arbitary dataset
print("Computing arbitary dataset ")
trained_models = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
trained_models = natsorted(trained_models)
model = trained_models[6]
model = model.split('/')
model = model[-1]
print('Load trained network: %s\n' % model)
dcgan.load(FLAGS.checkpoint_dir, model)
datapath = '/research3/datain/gmchoe_normal/0403/IR_0.25'
savepath = datapath
mean_nir = -0.3313
img_files = glob.glob(os.path.join(datapath, '*.png'))
img_files = natsorted(img_files)
pdb.set_trace()
for idx in xrange(0, len(img_files)):
print('Processing %d/%d \n' % (len(img_files), idx))
input_ = scipy.misc.imread(img_files[idx],
'F').astype(float)
height_size = input_.shape[0]
width_size = input_.shape[1]
input_ = np.reshape(
input_,
(height_size, width_size, 1)) # LF size:383 x 552
nondetail_input_ = ndimage.gaussian_filter(input_,
sigma=(1, 1, 0),
order=0)
input_ = input_ / 127.5 - 1.0
nondetail_input_ = nondetail_input_ / 127.5 - 1.0 # normalize -1 ~1
detail_input_ = input_ - nondetail_input_
nondetail_input_ = np.reshape(
nondetail_input_,
(1, height_size, width_size, 1)) # LF size:383 x 552
detail_input_ = np.reshape(detail_input_,
(1, height_size, width_size, 1))
#detail_input_ = detail_input_/127.5 -1.0 # normalize -1 ~1
start_time = time.time()
sample = sess.run(dcgan.G,
feed_dict={
dcgan.nondetail_images:
nondetail_input_,
dcgan.detail_images: detail_input_
})
print('time: %.8f' % (time.time() - start_time))
sample = np.squeeze(sample).astype(np.float32)
# normalization #
output = np.sqrt(np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output = (output + 1.) / 2.
"""
if not os.path.exists(os.path.join(savepath,'%s/%s/%s' %(FLAGS.dataset,model,listdir[idx]))):
os.makedirs(os.path.join(savepath,'%s/%s/%s' %(FLAGS.dataset,model,listdir[idx])))
"""
savename = os.path.join(
savepath, 'result/%s.bmp' % (img_files[idx][-10:]))
#savename = os.path.join(savepath,'single_normal_%02d.bmp' % (idx+1))
#savename = os.path.join(savepath,'%s/%s/%s/single_normal.bmp' % (FLAGS.dataset,model,listdir[idx]))
scipy.misc.imsave(savename, output)
elif VAL_OPTION == 2: # light source fixed
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
load, iteration = dcgan.load(FLAGS.checkpoint_dir)
pdb.set_trace()
savepath = './singleview_nir/L2ang/%d' % iteration
obj = 1
count = 1
if load:
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10): #tilt angles
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = './dataset/multi-view/testdata_3579/%03d/%03d/patch_%06d.mat' % (
obj, idx2, count)
input_ = scipy.io.loadmat(img)
input_ = input_['input_']
input_ = input_.astype(np.float)
#input_ = input_[:,:,0:3]
input_ = np.reshape(input_, (1, 600, 800, 4))
input_ = input_ / 127.5 - 1.0
start_time = time.time()
sample = sess.run([dcgan.G],
feed_dict={dcgan.images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.sqrt(
np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%03d/%03d' %
(list_val[idx], idx2))):
os.makedirs(
os.path.join(
savepath,
'%03d/%03d' % (list_val[idx], idx2)))
savename = os.path.join(
savepath, '%03d/%03d/multiview_normal.bmp' %
(list_val[idx], idx2))
scipy.misc.imsave(savename, output)
count = count + 1
obj = obj + 1
else:
print("Failed to load network")
def main(_):
pp.pprint(flags.FLAGS.__flags)
if not os.path.exists(FLAGS.checkpoint_dir):
os.makedirs(FLAGS.checkpoint_dir)
if not os.path.exists(FLAGS.sample_dir):
os.makedirs(FLAGS.sample_dir)
if not os.path.exists(os.path.join('./logs', time.strftime('%d%m'))):
os.makedirs(os.path.join('./logs', time.strftime('%d%m')))
gpu_config = tf.GPUOptions(per_process_gpu_memory_fraction=FLAGS.gpu)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_config)) as sess:
if FLAGS.is_train:
dcgan = DCGAN(sess, batch_size=FLAGS.batch_size,\
dataset_name=FLAGS.dataset, checkpoint_dir=FLAGS.checkpoint_dir)
else:
dcgan = EVAL(sess, batch_size=1,ir_image_shape=[256,256,1],normal_image_shape=[256,256,3],dataset_name=FLAGS.dataset,\
checkpoint_dir=FLAGS.checkpoint_dir)
print('deep model test \n')
if FLAGS.is_train:
dcgan.train(FLAGS)
else:
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
print '1: Estimating Normal maps from arbitary obejcts \n'
print '2: EStimating Normal maps according to only object tilt angles(Light direction is fixed(EX:3) \n'
print '3: Estimating Normal maps according to Light directions and object tilt angles \n'
x = input('Selecting a Evaluation mode:')
VAL_OPTION = int(x)
if VAL_OPTION == 1: # arbitary dataset
print("Computing arbitary dataset ")
trained_models = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
trained_models = natsorted(trained_models)
datapath = '/research2/Ammonight/*.bmp'
savepath = '/research2/Ammonight/output'
mean_nir = -0.3313
fulldatapath = os.path.join(glob.glob(datapath))
model = trained_models[4]
model = model.split('/')
model = model[-1]
pdb.set_trace()
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in xrange(len(fulldatapath)):
input_ = scipy.misc.imread(fulldatapath[idx]).astype(float)
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = (input_ / 127.5) - 1. # normalize -1 ~1
input_ = np.reshape(
input_, (1, input_.shape[0], input_.shape[1], 1))
input_ = np.array(input_).astype(np.float32)
mask = [input_ > -1.0][0] * 1.0
mean_mask = mask * mean_nir
#input_ = input_ - mean_mask
start_time = time.time()
sample = sess.run(dcgan.sampler,
feed_dict={dcgan.ir_images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.sqrt(np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output[output == inf] = 0.0
sample = (output + 1.0) / 2.0
name = fulldatapath[idx].split('/')
name = name[-1].split('.')
name = name[0]
savename = savepath + '/normal_' + name + '.bmp'
scipy.misc.imsave(savename, sample)
elif VAL_OPTION == 2: # light source fixed
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
save_files = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
save_files = natsorted(save_files)
savepath = './RMSS_ang_scale_loss_result'
for model_idx in range(0, len(save_files), 2):
model = save_files[model_idx]
model = model.split('/')
model = model[-1]
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10):
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(img +
'/3.bmp').astype(float)
gt_ = scipy.misc.imread(
'/research2/IR_normal_small/save016/1/12_Normal.bmp'
).astype(float)
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = (input_ / 127.5) - 1. # normalize -1 ~1
gt_ = scipy.misc.imresize(gt_, [600, 800])
gt_ = np.reshape(gt_, (1, 600, 800, 3))
gt_ = np.array(gt_).astype(np.float32)
input_ = np.reshape(input_, (1, 600, 800, 1))
input_ = np.array(input_).astype(np.float32)
start_time = time.time()
sample = sess.run(
dcgan.sampler,
feed_dict={dcgan.ir_images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.zeros((600, 800, 3)).astype(np.float32)
output[:, :, 0] = sample[:, :, 0] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 1] = sample[:, :, 1] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 2] = sample[:, :, 2] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[output == inf] = 0.0
sample = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%03d/%d' %
(list_val[idx], idx2))):
os.makedirs(
os.path.join(
savepath,
'%03d/%d' % (list_val[idx], idx2)))
savename = os.path.join(
savepath, '%03d/%d/single_normal_%s.bmp' %
(list_val[idx], idx2, model))
scipy.misc.imsave(savename, sample)
elif VAL_OPTION == 3: # depends on light sources
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
selec_model = [-2]
mean_nir = -0.3313 #-1~1
save_files = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
save_files = natsorted(save_files)
savepath = './Deconv_L1_result'
if not os.path.exists(os.path.join(savepath)):
os.makedirs(os.path.join(savepath))
for m in range(len(selec_model)):
model = save_files[selec_model[m]]
model = model.split('/')
model = model[-1]
print('Load model: %s \n' % model)
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10): #tilt angles 1~9
for idx3 in range(1, 13): # light source
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/ECCV_dataset_resized/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(
img + '/%d.bmp' % idx3).astype(
float) #input NIR image
input_ = scipy.misc.imresize(
input_, [256, 256])
input_ = input_ / 127.5 - 1.0 # normalize -1 ~1
input_ = np.reshape(input_, (1, 256, 256, 1))
input_ = np.array(input_).astype(np.float32)
gt_ = scipy.misc.imread(
img + '/12_Normal.bmp').astype(float)
gt_ = np.sum(gt_, axis=2)
gt_ = scipy.misc.imresize(gt_, [256, 256])
gt_ = np.reshape(gt_, [1, 256, 256, 1])
mask = [gt_ > 0.0][0] * 1.0
mean_mask = mean_nir * mask
#input_ = input_ - mean_mask
start_time = time.time()
sample = sess.run(dcgan.sampler,
feed_dict={
dcgan.ir_images: input_,
dcgan.keep_prob: 1.0
})
print('time: %.8f' %
(time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.sqrt(
np.sum(np.power(sample, 2), axis=2))
output = np.expand_dims(output, axis=-1)
output = sample / output
output = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%03d/%d/%s' %
(list_val[idx], idx2, model))):
os.makedirs(
os.path.join(
savepath, '%03d/%d/%s' %
(list_val[idx], idx2, model)))
savename = os.path.join(
savepath,
'%03d/%d/%s/single_normal_%03d.bmp' %
(list_val[idx], idx2, model, idx3))
scipy.misc.imsave(savename, output)
def main(_):
pp.pprint(flags.FLAGS.__flags)
if not os.path.exists(FLAGS.checkpoint_dir):
os.makedirs(FLAGS.checkpoint_dir)
if not os.path.exists(FLAGS.sample_dir):
os.makedirs(FLAGS.sample_dir)
if not os.path.exists(os.path.join('./logs', time.strftime('%d%m'))):
os.makedirs(os.path.join('./logs', time.strftime('%d%m')))
gpu_config = tf.GPUOptions(per_process_gpu_memory_fraction=FLAGS.gpu)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_config)) as sess:
if FLAGS.is_train:
if FLAGS.model == "narrow":
dcgan = DCGAN(sess, image_size=FLAGS.image_size, batch_size=FLAGS.batch_size,\
input_size=FLAGS.input_size,dataset_name=FLAGS.dataset,is_crop=FLAGS.is_crop, checkpoint_dir=FLAGS.checkpoint_dir)
else:
dcgan = Deep_DCGAN(sess, image_size=FLAGS.image_size, batch_size=FLAGS.batch_size,\
input_size=FLAGS.input_size,dataset_name=FLAGS.dataset,is_crop=FLAGS.is_crop, checkpoint_dir=FLAGS.checkpoint_dir)
else:
if FLAGS.model == "narrow":
dcgan = EVAL(sess, input_size = 600, batch_size=1,ir_image_shape=[None,None,1],normal_image_shape=[None,None,3],dataset_name=FLAGS.dataset,\
is_crop=False, checkpoint_dir=FLAGS.checkpoint_dir)
else:
dcgan = Deep_EVAL(sess, input_size = 600, batch_size=1,ir_image_shape=[None,None,1],normal_image_shape=[None,None,3],dataset_name=FLAGS.dataset,\
is_crop=False, checkpoint_dir=FLAGS.checkpoint_dir)
if FLAGS.is_train:
dcgan.train(FLAGS)
else:
OPTION = 2 # for validation
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
VAL_OPTION = 3
"""
if OPTION == 1:
data = json.load(open("/research2/IR_normal_small/json/traininput_single_224_ori_small.json"))
data_label = json.load(open("/research2/IR_normal_small/json/traingt_single_224_ori_small.json"))
elif OPTION == 2:
data = json.load(open("/research2/IR_normal_small/json/testinput_single_224_ori_small.json"))
data_label = json.load(open("/research2/IR_normal_small/json/testgt_single_224_ori_small.json"))
"""
if VAL_OPTION == 1:
model = 'DCGAN.model-10000'
dcgan.load(FLAGS.checkpoint_dir, model)
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
for idx in range(len(list_val)):
os.makedirs(
os.path.join('L1_loss_result', '%03d' % list_val[idx]))
for idx2 in range(1, 10):
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(img +
'/3.bmp').astype(float)
gt_ = scipy.misc.imread(
'/research2/IR_normal_small/save016/1/12_Normal.bmp'
).astype(float)
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = input_ / 255.0 - 1.0 # normalize -1 ~1
gt_ = scipy.misc.imresize(gt_, [600, 800])
#input_ = input_[240:840,515:1315]
#gt_ = gt_[240:840,515:1315]
input_ = np.reshape(input_, (1, 600, 800, 1))
gt_ = np.reshape(gt_, (1, 600, 800, 3))
input_ = np.array(input_).astype(np.float32)
gt_ = np.array(gt_).astype(np.float32)
start_time = time.time()
sample = sess.run(dcgan.sampler,
feed_dict={dcgan.ir_images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
gt_ = np.squeeze(gt_).astype(np.float32)
output = np.zeros((600, 800, 3)).astype(np.float32)
output[:, :, 0] = sample[:, :, 0] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 1] = sample[:, :, 1] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 2] = sample[:, :, 2] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[output == inf] = 0.0
sample = (output + 1.) / 2.
os.makedirs(
os.path.join('L1_loss_result',
'%03d/%d' % (list_val[idx], idx2)))
savename = './L1_loss_result/%03d/%d/single_normal_L1_%s.bmp' % (
list_val[idx], idx2, model)
scipy.misc.imsave(savename, sample)
elif VAL_OPTION == 2: # arbitary dataset
print("Computing arbitary dataset ")
trained_models = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
trained_models = natsorted(trained_models)
datapath = '/home/yjyoon/Dropbox/ECCV_result/smartphone/iphone/input/gray_*.bmp'
savepath = '/home/yjyoon/Dropbox/ECCV_result/smartphone/iphone/output'
fulldatapath = os.path.join(glob.glob(datapath))
model = trained_models[-2]
model = model.split('/')
model = model[-1]
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in xrange(len(fulldatapath)):
#input_ = cv2.imread(fulldatapath[idx])
#input_ = cv2.cvtColor(input_,cv2.COLOR_BGR2YCR_CB)
#input_ = cv2.resize(input_[:,:,0],(600,800))
input_ = scipy.misc.imread(fulldatapath[idx]).astype(float)
input_ = (input_ / 127.5) - 1. # normalize -1 ~1
input_ = np.reshape(
input_, (1, input_.shape[0], input_.shape[1], 1))
#[Y,Cr,Cb]= rgb2ycbcr(input_)
#input_= rgb2gray(input_)
#input_ = scipy.misc.imresize(Y,(600,800))
#input_ = np.reshape(input_,(1,600,800,1))
input_ = np.array(input_).astype(np.float32)
start_time = time.time()
sample = sess.run(dcgan.sampler,
feed_dict={dcgan.ir_images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.zeros((sample.shape[0], sample.shape[1],
3)).astype(np.float32)
output[:, :, 0] = sample[:, :, 0] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 1] = sample[:, :, 1] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 2] = sample[:, :, 2] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[output == inf] = 0.0
sample = (output + 1.) / 2.
name = fulldatapath[idx].split('/')
name = name[-1].split('.')
name = name[0]
savename = savepath + '/normal_' + name + '.bmp'
scipy.misc.imsave(savename, sample)
elif VAL_OPTION == 3: # light source fixed
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
save_files = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
save_files = natsorted(save_files)
savepath = './RMSS_ang_scale_loss_result'
max_h = 600
max_w = 800
for model_idx in range(0, len(save_files), 2):
model = save_files[model_idx]
model = model.split('/')
model = model[-1]
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath,
'%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10):
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(img +
'/3.bmp').astype(float)
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = (input_ / 127.5) - 1. # normalize -1 ~1
overlap = np.min(max_h, max_w) - 100
step_h = np.ceil(max_h, overlap)
step_w = np.ceil(max_w, overlap)
result = np.zeros(input_.shape[0], input_.shape[1],
input_.shape[2], step_h * step_w)
tmp_result = []
for h in range(0, input_.shape[0], overlap):
for w in range(0, input_.shape[1], overlap):
crop_input_ = input_[h:h + max_h,
w:w + max_w]
crop_input_ = np.reshape(
crop_input_, (max_h, max_w, 1))
crop_intpu_ = np.array(crop_input_).astype(
np.float32)
start_time = time.time()
sample = sess.run(dcgan.sampler,
feed_dict={
dcgan.ir_images:
crop_input_
})
print('time: %.8f' %
(time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(
np.float32)
output = np.sqrt(
np.sum(np.square(sample), axis=2))
output = np.expand_dims(output, -1)
output[output == 0.0] = 1e-10
output = sample / output
tmp_result.append(output)
result = recovering_fullimage(
tmp_result, result, overlap, max_h, max_w)
if not os.path.exists(
os.path.join(
savepath, '%03d/%d' %
(list_val[idx], idx2))):
os.makedirs(
os.path.join(
savepath,
'%03d/%d' % (list_val[idx], idx2)))
savename = os.path.join(
savepath, '%03d/%d/single_normal_%s.bmp' %
(list_val[idx], idx2, model))
scipy.misc.imsave(savename, result)
elif VAL_OPTION == 4: # depends on light sources
list_val = [11, 16, 21, 22, 33, 36, 38, 53, 59, 92]
save_files = glob.glob(
os.path.join(FLAGS.checkpoint_dir, FLAGS.dataset,
'DCGAN.model*'))
save_files = natsorted(save_files)
savepath = './L1_ang_loss_lights_result'
if not os.path.exists(os.path.join(savepath)):
os.makedirs(os.path.join(savepath))
model = save_files[-2]
model = model.split('/')
model = model[-1]
dcgan.load(FLAGS.checkpoint_dir, model)
for idx in range(len(list_val)):
if not os.path.exists(
os.path.join(savepath, '%03d' % list_val[idx])):
os.makedirs(
os.path.join(savepath, '%03d' % list_val[idx]))
for idx2 in range(1, 10): #tilt angles 1~9
for idx3 in range(1, 13): # light source
print("Selected material %03d/%d" %
(list_val[idx], idx2))
img = '/research2/IR_normal_small/save%03d/%d' % (
list_val[idx], idx2)
input_ = scipy.misc.imread(
img + '/%d.bmp' % idx3).astype(
float) #input NIR image
input_ = scipy.misc.imresize(input_, [600, 800])
input_ = input_ / 127.5 - 1.0 # normalize -1 ~1
input_ = np.reshape(input_, (1, 600, 800, 1))
input_ = np.array(input_).astype(np.float32)
start_time = time.time()
sample = sess.run(
dcgan.sampler,
feed_dict={dcgan.ir_images: input_})
print('time: %.8f' % (time.time() - start_time))
# normalization #
sample = np.squeeze(sample).astype(np.float32)
output = np.zeros((600, 800, 3)).astype(np.float32)
output[:, :, 0] = sample[:, :, 0] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 1] = sample[:, :, 1] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[:, :, 2] = sample[:, :, 2] / (np.sqrt(
np.power(sample[:, :, 0], 2) +
np.power(sample[:, :, 1], 2) +
np.power(sample[:, :, 2], 2)))
output[output == inf] = 0.0
sample = (output + 1.) / 2.
if not os.path.exists(
os.path.join(
savepath, '%03d/%d' %
(list_val[idx], idx2))):
os.makedirs(
os.path.join(
savepath,
'%03d/%d' % (list_val[idx], idx2)))
savename = os.path.join(
savepath, '%03d/%d/single_normal_%d.bmp' %
(list_val[idx], idx2, idx3))
scipy.misc.imsave(savename, sample)
def recovering_fullimage(output, result, overlap, max_h, max_w):
count = 0
for h in range(0, input_.shape[0], overlap):
for w in range(0, input_.shape[1], overlap):
result[h:h + max_h, w:w + max_w, :, count] = output[count]
count += 0
return np.max(result, axis=3)