# Copyright UCL Business plc 2017. Patent Pending. All rights reserved.
#
# The MonoDepth Software is licensed under the terms of the UCLB ACP-A licence
# which allows for non-commercial use only, the full terms of which are made
# available in the LICENSE file.
#
# For any other use of the software not covered by the UCLB ACP-A Licence,
# please contact info@uclb.com
"""Fully convolutional model for monocular depth estimation
by Clement Godard, Oisin Mac Aodha and Gabriel J. Brostow
http://visual.cs.ucl.ac.uk/pubs/monoDepth/
"""
from __future__ import absolute_import, division, print_function
from collections import namedtuple
import numpy as np
import tensorflow as tf
import tensorflow.contrib.slim as slim
from utils import *
from bilinear_sampler import *
monodepth_parameters = namedtuple('parameters',
'encoder, '
'height, width, '
'batch_size, '
'num_threads, '
'num_epochs, '
'do_stereo, '
'wrap_mode, '
'use_deconv, '
'alpha_image_loss, '
'disp_gradient_loss_weight, '
'lr_loss_weight, '
'task, '
'full_summary')
class MonodepthModel(object):
"""monodepth model"""
def __init__(self, params, mode, task, left, right, semantic, valid, reuse_variables=None, model_index=0):
self.params = params
self.mode = mode
self.left = left
self.right = right
self.semantic = semantic
self.valid = valid
self.model_collection = ['model_' + str(model_index)]
self.task = task
self.classes = get_num_classes()
self.reuse_variables = reuse_variables
self.build_model()
self.build_outputs()
if self.mode == 'test':
return
self.build_losses()
self.build_summaries()
def gradient_x(self, img):
gx = img[:,:,:-1,:] - img[:,:,1:,:]
return gx
def gradient_y(self, img):
gy = img[:,:-1,:,:] - img[:,1:,:,:]
return gy
def averaged_gradient_x(self, img):
gx = (img[:,:,:-3,:] - img[:,:,3:,:] + img[:,:,:-3,:] - img[:,:,2:-1,:] + img[:,:,:-3,:] - img[:,:,1:-2,:]) / 3.
return gx
def averaged_gradient_y(self, img):
gy = (img[:,:-3,:,:] - img[:,3:,:,:] + img[:,:-3,:,:] - img[:,2:-1,:,:] + img[:,:-3,:,:] - img[:,1:-2,:,:]) / 3.
return gy
def upsample_nn(self, x, ratio):
s = tf.shape(x)
h = s[1]
w = s[2]
return tf.image.resize_nearest_neighbor(x, [h * ratio, w * ratio])
def scale_pyramid(self, img, num_scales):
scaled_imgs = [img]
s = tf.shape(img)
h = s[1]
w = s[2]
for i in range(num_scales - 1):
ratio = 2 ** (i + 1)
nh = h // ratio
nw = w // ratio
scaled_imgs.append(tf.image.resize_area(img, [nh, nw]))
return scaled_imgs
def generate_image_left(self, img, disp):
return bilinear_sampler_1d_h(img, -disp)
def generate_image_right(self, img, disp):
return bilinear_sampler_1d_h(img, disp)
def SSIM(self, x, y):
C1 = 0.01 ** 2
C2 = 0.03 ** 2
mu_x = slim.avg_pool2d(x, 3, 1, 'VALID')
mu_y = slim.avg_pool2d(y, 3, 1, 'VALID')
sigma_x = slim.avg_pool2d(x ** 2, 3, 1, 'VALID') - mu_x ** 2
sigma_y = slim.avg_pool2d(y ** 2, 3, 1, 'VALID') - mu_y ** 2
sigma_xy = slim.avg_pool2d(x * y , 3, 1, 'VALID') - mu_x * mu_y
SSIM_n = (2 * mu_x * mu_y + C1) * (2 * sigma_xy + C2)
SSIM_d = (mu_x ** 2 + mu_y ** 2 + C1) * (sigma_x + sigma_y + C2)
SSIM = SSIM_n / SSIM_d
return tf.clip_by_value((1 - SSIM) / 2, 0, 1)
def get_disparity_smoothness(self, disp, pyramid):
disp_gradients_x = [self.gradient_x(d) for d in disp]
disp_gradients_y = [self.gradient_y(d) for d in disp]
image_gradients_x = [self.gradient_x(img) for img in pyramid]
image_gradients_y = [self.gradient_y(img) for img in pyramid]
weights_x = [tf.exp(-tf.reduce_mean(tf.abs(g), 3, keep_dims=True)) for g in image_gradients_x]
weights_y = [tf.exp(-tf.reduce_mean(tf.abs(g), 3, keep_dims=True)) for g in image_gradients_y]
smoothness_x = [disp_gradients_x[i] * weights_x[i] for i in range(4)]
smoothness_y = [disp_gradients_y[i] * weights_y[i] for i in range(4)]
return smoothness_x + smoothness_y
def get_disparity_edge(self, disp, semantic):
# gradient normalized by disparity
disp_gradients_x = tf.divide(self.gradient_x(disp), disp[:,:,:-1,:]+10e-10)
disp_gradients_y = tf.divide(self.gradient_y(disp), disp[:,:-1,:,:]+10e-10)
self.semantic_gradients_x = tf.cast(tf.sign(tf.abs(self.gradient_x(semantic))), tf.float32)
self.semantic_gradients_y = tf.cast(tf.sign(tf.abs(self.gradient_y(semantic))), tf.float32)
# AVERAGED GRADIENTS (DISTANCE 1, 2, 3)
# disp_gradients_x = self.averaged_gradient_x(disp)
# disp_gradients_y = self.averaged_gradient_y(disp)
# self.semantic_gradients_x = tf.sign(tf.abs(self.averaged_gradient_x(tf.cast(semantic, tf.float32))))
# self.semantic_gradients_y = tf.sign(tf.abs(self.averaged_gradient_y(tf.cast(semantic, tf.float32))))
weights_x = tf.exp(-tf.abs(disp_gradients_x))
weights_y = tf.exp(-tf.abs(disp_gradients_y))
edge_x = self.semantic_gradients_x * weights_x
edge_y = self.semantic_gradients_y * weights_y
return edge_x, edge_y
def get_disp(self, x):
disp = 0.3 * self.conv(x, 2, 3, 1, tf.nn.sigmoid)
return disp
def conv(self, x, num_out_layers, kernel_size, stride, activation_fn=tf.nn.elu):
p = np.floor((kernel_size - 1) / 2).astype(np.int32)
p_x = tf.pad(x, [[0, 0], [p, p], [p, p], [0, 0]])
return slim.conv2d(p_x, num_out_layers, kernel_size, stride, 'VALID', activation_fn=activation_fn)
def dilated_conv(self, x, num_out_layers, kernel_size, stride, activation_fn=tf.nn.elu, dilation_rate=1):
p = np.floor((kernel_size - 1) / 2).astype(np.int32)
p_x = tf.pad(x, [[0, 0], [p, p], [p, p], [0, 0]])
return slim.conv2d(p_x, num_out_layers, kernel_size, stride, 'VALID', activation_fn=activation_fn, rate=dilation_rate)
def conv_block(self, x, num_out_layers, kernel_size):
conv1 = self.conv(x, num_out_layers, kernel_size, 1)
conv2 = self.conv(conv1, num_out_layers, kernel_size, 2)
return conv2
def maxpool(self, x, kernel_size):
p = np.floor((kernel_size - 1) / 2).astype(np.int32)
p_x = tf.pad(x, [[0, 0], [p, p], [p, p], [0, 0]])
return slim.max_pool2d(p_x, kernel_size)
def resconv(self, x, num_layers, stride):
do_proj = tf.shape(x)[3] != num_layers or stride == 2
shortcut = []
conv1 = self.conv(x, num_layers, 1, 1)
conv2 = self.conv(conv1, num_layers, 3, stride)
conv3 = self.conv(conv2, 4 * num_layers, 1, 1, None)
if do_proj:
shortcut = self.conv(x, 4 * num_layers, 1, stride, None)
else:
shortcut = x
return tf.nn.elu(conv3 + shortcut)
def resblock(self, x, num_layers, num_blocks):
out = x
for i in range(num_blocks - 1):
out = self.resconv(out, num_layers, 1)
out = self.resconv(out, num_layers, 2)
return out
def upconv(self, x, num_out_layers, kernel_size, scale):
upsample = self.upsample_nn(x, scale)
conv = self.conv(upsample, num_out_layers, kernel_size, 1)
return conv
def deconv(self, x, num_out_layers, kernel_size, scale):
p_x = tf.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]])
conv = slim.conv2d_transpose(p_x, num_out_layers, kernel_size, scale, 'SAME')
return conv[:,3:-1,3:-1,:]
def build_vgg(self,inputs):
#set convenience functions
conv = self.conv
if self.params.use_deconv:
upconv = self.deconv
else:
upconv = self.upconv
with tf.variable_scope('encoder'):
conv1 = self.conv_block(inputs, 32, 7) # H/2
conv2 = self.conv_block(conv1, 64, 5) # H/4
conv3 = self.conv_block(conv2, 128, 3) # H/8
conv4 = self.conv_block(conv3, 256, 3) # H/16
conv5 = self.conv_block(conv4, 512, 3) # H/32
conv6 = self.conv_block(conv5, 512, 3) # H/64
conv7 = self.conv_block(conv6, 512, 3) # H/128
with tf.variable_scope('skips'):
skip1 = conv1
skip2 = conv2
skip3 = conv3
skip4 = conv4
skip5 = conv5
skip6 = conv6
if 'depth' in self.task:
with tf.variable_scope('decoder'):
upconv7_d = upconv(conv7, 512, 3, 2) #H/64
concat7_d = tf.concat([upconv7_d, skip6], 3)
iconv7_d = conv(concat7_d, 512, 3, 1)
upconv6_d = upconv(iconv7_d, 512, 3, 2) #H/32
concat6_d = tf.concat([upconv6_d, skip5], 3)
iconv6_d = conv(concat6_d, 512, 3, 1)
upconv5_d = upconv(iconv6_d, 256, 3, 2) #H/16
concat5_d = tf.concat([upconv5_d, skip4], 3)
iconv5_d = conv(concat5_d, 256, 3, 1)
upconv4_d = upconv(iconv5_d, 128, 3, 2) #H/8
concat4_d = tf.concat([upconv4_d, skip3], 3)
iconv4_d = conv(concat4_d, 128, 3, 1)
self.disp4 = self.get_disp(iconv4_d)
udisp4_d = self.upsample_nn(self.disp4, 2)
upconv3_d = upconv(iconv4_d, 64, 3, 2) #H/4
concat3_d = tf.concat([upconv3_d, skip2, udisp4_d], 3)
iconv3_d = conv(concat3_d, 64, 3, 1)
self.disp3 = self.get_disp(iconv3_d)
udisp3_d = self.upsample_nn(self.disp3, 2)
upconv2_d = upconv(iconv3_d, 32, 3, 2) #H/2
concat2_d = tf.concat([upconv2_d, skip1, udisp3_d], 3)
iconv2_d = conv(concat2_d, 32, 3, 1)
self.disp2 = self.get_disp(iconv2_d)
udisp2_d = self.upsample_nn(self.disp2, 2)
upconv1_d = upconv(iconv2_d, 16, 3, 2) #H
concat1_d = tf.concat([upconv1_d, udisp2_d], 3)
iconv1_d = conv(concat1_d, 16, 3, 1)
self.disp1 = self.get_disp(iconv1_d)
if 'semantic' in self.task:
with tf.variable_scope('decoder-semantic'):
upconv7_s = upconv(conv7, 512, 3, 2) #H/64
concat7_s = tf.concat([upconv7_s, skip6], 3)
iconv7_s = conv(concat7_s, 512, 3, 1)
upconv6_s = upconv(iconv7_s, 512, 3, 2) #H/32
concat6_s = tf.concat([upconv6_s, skip5], 3)
iconv6_s = conv(concat6_s, 512, 3, 1)
upconv5_s = upconv(iconv6_s, 256, 3, 2) #H/16
concat5_s = tf.concat([upconv5_s, skip4], 3)
iconv5_s = conv(concat5_s, 256, 3, 1)
upconv4_s = upconv(iconv5_s, 128, 3, 2) #H/8
concat4_s = tf.concat([upconv4_s, skip3], 3)
iconv4_s = conv(concat4_s, 128, 3, 1)
uconv4_s = self.upsample_nn(iconv4_s, 2)
upconv3_s = upconv(iconv4_s, 64, 3, 2) #H/4
concat3_s = tf.concat([upconv3_s, skip2, uconv4_s], 3)
iconv3_s = conv(concat3_s, 64, 3, 1)
uconv3_s = self.upsample_nn(iconv3_s, 2)
upconv2_s = upconv(iconv3_s, 32, 3, 2) #H/2
concat2_s = tf.concat([upconv2_s, skip1, uconv3_s], 3)
iconv2_s = conv(concat2_s, 32, 3, 1)
uconv2_s = self.upsample_nn(iconv2_s, 2)
upconv1_s = upconv(iconv2_s, 32, 3, 2) #H
concat1_s = tf.concat([upconv1_s, uconv2_s], 3)
iconv1_s = conv(concat1_s, self.classes, 3, 1)
self.sem1 = iconv1_s
def build_resnet50(self):
#set convenience functions
conv = self.conv
if self.params.use_deconv:
upconv = self.deconv
else:
upconv = self.upconv
with tf.variable_scope('encoder'):
conv1 = conv(self.model_input, 64, 7, 2) # H/2 - 64D
pool1 = self.maxpool(conv1, 3) # H/4 - 64D
conv2 = self.resblock(pool1, 64, 3) # H/8 - 256D
conv3 = self.resblock(conv2, 128, 4) # H/16 - 512D
conv4 = self.resblock(conv3, 256, 6) # H/32 - 1024D
conv5 = self.resblock(conv4, 512, 3) # H/64 - 2048D
with tf.variable_scope('skips'):
skip1 = conv1
skip2 = pool1
skip3 = conv2
skip4 = conv3
skip5 = conv4
if 'depth' in self.task:
with tf.variable_scope('decoder'):
upconv6_d = upconv(conv5, 512, 3, 2) #H/32
concat6_d = tf.concat([upconv6_d, skip5], 3)
iconv6_d = conv(concat6_d, 512, 3, 1)
upconv5_d = upconv(iconv6_d, 256, 3, 2) #H/16
concat5_d = tf.concat([upconv5_d, skip4], 3)
iconv5_d = conv(concat5_d, 256, 3, 1)
upconv4_d = upconv(iconv5_d, 128, 3, 2) #H/8
concat4_d = tf.concat([upconv4_d, skip3], 3)
iconv4_d = conv(concat4_d, 128, 3, 1)
self.disp4 = self.get_disp(iconv4_d)
udisp4_d = self.upsample_nn(self.disp4, 2)
upconv3_d = upconv(iconv4_d, 64, 3, 2) #H/4
concat3_d = tf.concat([upconv3_d, skip2, udisp4_d], 3)
iconv3_d = conv(concat3_d, 64, 3, 1)
self.disp3 = self.get_disp(iconv3_d)
udisp3_d = self.upsample_nn(self.disp3, 2)
upconv2_d = upconv(iconv3_d, 32, 3, 2) #H/2
concat2_d = tf.concat([upconv2_d, skip1, udisp3_d], 3)
iconv2_d = conv(concat2_d, 32, 3, 1)
self.disp2 = self.get_disp(iconv2_d)
udisp2_d = self.upsample_nn(self.disp2, 2)
upconv1_d = upconv(iconv2_d, 16, 3, 2) #H
concat1_d = tf.concat([upconv1_d, udisp2_d], 3)
iconv1_d = conv(concat1_d, 16, 3, 1)
self.disp1 = self.get_disp(iconv1_d)
if 'semantic' in self.task:
with tf.variable_scope('decoder-semantic'):
upconv6_s = upconv(conv5, 512, 3, 2) #H/32
concat6_s = tf.concat([upconv6_s, skip5], 3)
iconv6_s = conv(concat6_s, 512, 3, 1)
upconv5_s = upconv(iconv6_s, 256, 3, 2) #H/16
concat5_s = tf.concat([upconv5_s, skip4], 3)
iconv5_s = conv(concat5_s, 256, 3, 1)
upconv4_s = upconv(iconv5_s, 128, 3, 2) #H/8
concat4_s = tf.concat([upconv4_s, skip3], 3)
iconv4_s = conv(concat4_s, 128, 3, 1)
uconv4_s = self.upsample_nn(iconv4_s, 2)
upconv3_s = upconv(iconv4_s, 64, 3, 2) #H/4
concat3_s = tf.concat([upconv3_s, skip2, uconv4_s], 3)
iconv3_s = conv(concat3_s, 64, 3, 1)
uconv3_s = self.upsample_nn(iconv3_s, 2)
upconv2_s = upconv(iconv3_s, 32, 3, 2) #H/2
concat2_s = tf.concat([upconv2_s, skip1, uconv3_s], 3)
iconv2_s = conv(concat2_s, 32, 3, 1)
uconv2_s = self.upsample_nn(iconv2_s, 2)
upconv1_s = upconv(iconv2_s, 32, 3, 2) #H
concat1_s = tf.concat([upconv1_s, uconv2_s], 3)
iconv1_s = conv(concat1_s, self.classes, 3, 1)
self.sem1 = iconv1_s
def build_model(self):
with slim.arg_scope([slim.conv2d, slim.conv2d_transpose], activation_fn=tf.nn.elu):
with tf.variable_scope('model', reuse=self.reuse_variables):
self.left_pyramid = self.scale_pyramid(self.left, 4)
if self.mode == 'train':
self.right_pyramid = self.scale_pyramid(self.right, 4)
self.model_input = self.left
#build model
if self.params.encoder == 'vgg':
self.build_vgg(self.model_input)
elif self.params.encoder == 'resnet50':
self.build_resnet50()
else:
return None
def build_outputs(self):
# STORE DISPARITIES
if 'depth' in self.task:
with tf.variable_scope('disparities'):
self.disp_est = [self.disp1, self.disp2, self.disp3, self.disp4]
self.disp_left_est = [tf.expand_dims(d[:,:,:,0], 3) for d in self.disp_est]
self.disp_right_est = [tf.expand_dims(d[:,:,:,1], 3) for d in self.disp_est]
if 'semantic' in self.task:
with tf.variable_scope('semantic'):
self.sem_est = [self.sem1] # will we include multi-scale semantic?
if self.mode == 'test':
return
# GENERATE IMAGES
if 'depth' in self.task:
with tf.variable_scope('images'):
self.left_est = [self.generate_image_left(self.right_pyramid[i], self.disp_left_est[i]) for i in range(4)]
self.right_est = [self.generate_image_right(self.left_pyramid[i], self.disp_right_est[i]) for i in range(4)]
# LR CONSISTENCY
with tf.variable_scope('left-right'):
self.right_to_left_disp = [self.generate_image_left(self.disp_right_est[i], self.disp_left_est[i]) for i in range(4)]
self.left_to_right_disp = [self.generate_image_right(self.disp_left_est[i], self.disp_right_est[i]) for i in range(4)]
# DISPARITY SMOOTHNESS
with tf.variable_scope('smoothness'):
self.disp_left_smoothness = self.get_disparity_smoothness(self.disp_left_est, self.left_pyramid)
self.disp_right_smoothness = self.get_disparity_smoothness(self.disp_right_est, self.right_pyramid)
def build_losses(self):
with tf.variable_scope('losses', reuse=self.reuse_variables):
self.total_loss = 0
if 'depth' in self.task:
# IMAGE RECONSTRUCTION
# L1
self.l1_left = [tf.abs( self.left_est[i] - self.left_pyramid[i]) for i in range(4)]
self.l1_reconstruction_loss_left = [tf.reduce_mean(l) for l in self.l1_left]
self.l1_right = [tf.abs(self.right_est[i] - self.right_pyramid[i]) for i in range(4)]
self.l1_reconstruction_loss_right = [tf.reduce_mean(l) for l in self.l1_right]
# SSIM
self.ssim_left = [self.SSIM( self.left_est[i], self.left_pyramid[i]) for i in range(4)]
self.ssim_loss_left = [tf.reduce_mean(s) for s in self.ssim_left]
self.ssim_right = [self.SSIM(self.right_est[i], self.right_pyramid[i]) for i in range(4)]
self.ssim_loss_right = [tf.reduce_mean(s) for s in self.ssim_right]
# WEIGTHED SUM
self.image_loss_right = [self.params.alpha_image_loss * self.ssim_loss_right[i] + (1 - self.params.alpha_image_loss) * self.l1_reconstruction_loss_right[i] for i in range(4)]
self.image_loss_left = [self.params.alpha_image_loss * self.ssim_loss_left[i] + (1 - self.params.alpha_image_loss) * self.l1_reconstruction_loss_left[i] for i in range(4)]
self.image_loss = tf.add_n(self.image_loss_left + self.image_loss_right)
# DISPARITY SMOOTHNESS
self.disp_left_loss = [tf.reduce_mean(tf.abs(self.disp_left_smoothness[i])) / 2 ** i for i in range(4)]
self.disp_right_loss = [tf.reduce_mean(tf.abs(self.disp_right_smoothness[i])) / 2 ** i for i in range(4)]
self.disp_gradient_loss = tf.add_n(self.disp_left_loss + self.disp_right_loss)
# LR CONSISTENCY
self.lr_left_loss = [tf.reduce_mean(tf.abs(self.right_to_left_disp[i] - self.disp_left_est[i])) for i in range(4)]
self.lr_right_loss = [tf.reduce_mean(tf.abs(self.left_to_right_disp[i] - self.disp_right_est[i])) for i in range(4)]
self.lr_loss = tf.add_n(self.lr_left_loss + self.lr_right_loss)
# TOTAL LOSS
self.total_loss += self.image_loss + self.params.disp_gradient_loss_weight * self.disp_gradient_loss + self.params.lr_loss_weight * self.lr_loss
# SEMANTIC LOSS
if 'semantic' in self.task:
self.semantic_loss = tf.reduce_mean(tf.multiply(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=self.semantic, logits=self.sem_est), self.valid)) * 0.1
self.total_loss += self.semantic_loss
if 'warp-semantic' in self.task:
self.semw = self.generate_image_left(self.sem2, self.disp_left_est[0])
self.semantic_loss_warp = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=self.semantic, logits=[self.semw])) * 0.1
if 'edge' in self.task:
self.disparity_edge = self.get_disparity_edge(self.disp1, self.semantic)
self.disp_edge_loss = tf.add_n([tf.reduce_mean(self.disparity_edge[i]) for i in range(len(self.disparity_edge))])
self.total_loss += 0.1*self.disp_edge_loss
def build_summaries(self):
# SUMMARIES
with tf.device('/cpu:0'):
tf.summary.image('left', self.left, max_outputs=4, collections=self.model_collection)
if 'semantic' in self.task:
tf.summary.image('pred_sem', colormap_semantic(tf.expand_dims(tf.argmax(self.sem1, axis=-1), axis=-1)), collections=self.model_collection)
tf.summary.image('gt_sem', colormap_semantic(self.semantic), collections=self.model_collection)
tf.summary.scalar('semantic_loss', self.semantic_loss, collections=self.model_collection)
if 'warp' in self.task:
tf.summary.image('pred_sem_right_warped', colormap_semantic(tf.expand_dims(tf.argmax(self.semw, axis=-1), axis=-1)), collections=self.model_collection)
if 'edge' in self.task:
tf.summary.scalar('disp_edge_loss', self.disp_edge_loss, collections=self.model_collection)
if 'depth' in self.task:
for i in range(4):
tf.summary.scalar('ssim_loss_' + str(i), self.ssim_loss_left[i] + self.ssim_loss_right[i], collections=self.model_collection)
tf.summary.scalar('l1_loss_' + str(i), self.l1_reconstruction_loss_left[i] + self.l1_reconstruction_loss_right[i], collections=self.model_collection)
tf.summary.scalar('image_loss_' + str(i), self.image_loss_left[i] + self.image_loss_right[i], collections=self.model_collection)
tf.summary.scalar('disp_gradient_loss_' + str(i), self.disp_left_loss[i] + self.disp_right_loss[i], collections=self.model_collection)
tf.summary.scalar('lr_loss_' + str(i), self.lr_left_loss[i] + self.lr_right_loss[i], collections=self.model_collection)
tf.summary.image('disp_left_est_' + str(i), self.disp_left_est[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('disp_right_est_' + str(i), self.disp_right_est[i], max_outputs=4, collections=self.model_collection)
if self.params.full_summary:
tf.summary.image('left_est_' + str(i), self.left_est[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('right_est_' + str(i), self.right_est[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('ssim_left_' + str(i), self.ssim_left[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('ssim_right_' + str(i), self.ssim_right[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('l1_left_' + str(i), self.l1_left[i], max_outputs=4, collections=self.model_collection)
tf.summary.image('l1_right_' + str(i), self.l1_right[i], max_outputs=4, collections=self.model_collection)
if self.params.full_summary:
tf.summary.image('right', self.right, max_outputs=4, collections=self.model_collection)
