def __getitem__(self, idx):
input_data = {}
# left
left_fname = self.left_data[idx]
left = np.array(Image.open(left_fname)).astype(np.uint8)
input_data['left'] = left
# right
right_fname = self.right_data[idx]
right = np.array(Image.open(right_fname)).astype(np.uint8)
input_data['right'] = right
# disp
if not self.split == 'test': # no disp for test files
disp_fname = self.disp_data[idx]
disp = np.array(Image.open(disp_fname)).astype(np.float) / 256.
input_data['disp'] = disp
input_data['occ_mask'] = np.zeros_like(disp).astype(np.bool)
if self.split == 'train':
input_data = random_crop(200, 640, input_data, self.split)
input_data = augment(input_data, self.transformation)
else:
input_data = normalization(**input_data)
return input_data
def __getitem__(self, idx):
input_data = {}
# left
left_fname = os.path.join(self.datadir, self.left_data[idx])
left = np.array(Image.open(left_fname)).astype(np.uint8)
input_data['left'] = left
# right
right_fname = os.path.join(self.datadir, self.right_data[idx])
right = np.array(Image.open(right_fname)).astype(np.uint8)
input_data['right'] = right
if not self.split == 'test': # no disp for test files
# occ
occ_left_fname = os.path.join(self.datadir,
self.occ_left_data[idx])
occ_right_fname = os.path.join(self.datadir,
self.occ_right_data[idx])
occ_left = np.array(Image.open(occ_left_fname)) == 128
occ_right = np.array(Image.open(occ_right_fname)) == 128
# disp
disp_left_fname = os.path.join(self.datadir,
self.disp_left_data[idx])
disp_right_fname = os.path.join(self.datadir,
self.disp_right_data[idx])
disp_left, _ = readPFM(disp_left_fname)
disp_right, _ = readPFM(disp_right_fname)
if self.split == 'train':
# horizontal flip
input_data['left'], input_data['right'], input_data['occ_mask'], input_data['occ_mask_right'], \
input_data['disp'], \
input_data['disp_right'] = horizontal_flip(input_data['left'], input_data['right'], occ_left, occ_right,
disp_left,
disp_right, self.split)
# random crop
input_data = random_crop(360, 640, input_data, self.split)
else:
input_data['occ_mask'] = occ_left
input_data['occ_mask_right'] = occ_right
input_data['disp'] = disp_left
input_data['disp_right'] = disp_right
input_data = augment(input_data, self.transformation)
else:
input_data = normalization(**input_data)
return input_data
def __getitem__(self, idx):
result = {}
left_fname = self.left_data[idx]
result['left'] = np.array(Image.open(left_fname)).astype(
np.uint8)[..., :3]
right_fname = left_fname.replace('left', 'right')
result['right'] = np.array(Image.open(right_fname)).astype(
np.uint8)[..., :3]
occ_right_fname = self.occ_data[idx].replace('left', 'right')
occ_left = np.array(Image.open(self.occ_data[idx])).astype(np.bool)
occ_right = np.array(Image.open(occ_right_fname)).astype(np.bool)
disp_left_fname = left_fname.replace('frame_finalpass',
'disparity').replace(
'.png', '.pfm')
disp_right_fname = right_fname.replace('frame_finalpass',
'disparity').replace(
'.png', '.pfm')
disp_left, _ = readPFM(disp_left_fname)
disp_right, _ = readPFM(disp_right_fname)
if self.split == "train":
# horizontal flip
result['left'], result['right'], result['occ_mask'], result['occ_mask_right'], disp, disp_right \
= horizontal_flip(result['left'], result['right'], occ_left, occ_right, disp_left, disp_right,
self.split)
result['disp'] = np.nan_to_num(disp, nan=0.0)
result['disp_right'] = np.nan_to_num(disp_right, nan=0.0)
# random crop
result = random_crop(360, 640, result, self.split)
else:
result['occ_mask'] = occ_left
result['occ_mask_right'] = occ_right
result['disp'] = disp_left
result['disp_right'] = disp_right
result = augment(result, self.transformation)
return result
def __getitem__(self, idx):
result = {}
left_fname = self.left_data[idx]
result['left'] = np.array(Image.open(left_fname)).astype(
np.uint8)[..., :3]
right_fname = left_fname.replace('left', 'right')
result['right'] = np.array(Image.open(right_fname)).astype(
np.uint8)[..., :3]
occ_left_fname = left_fname.replace("image_clean",
"disparity_occlusions")
occ_right_fname = right_fname.replace("image_clean",
"disparity_occlusions")
#occ_right_fname = self.occ_data[idx].replace('left', 'right')
occ_left = np.array(Image.open(occ_left_fname)).astype(np.bool)
occ_right = np.array(Image.open(occ_right_fname)).astype(np.bool)
disp_left_fname = left_fname.replace('image_clean',
'disparity').replace(
'.png', '.pfm')
disp_right_fname = right_fname.replace('image_clean',
'disparity').replace(
'.png', '.pfm')
disp_left, _ = readPFM(disp_left_fname)
disp_right, _ = readPFM(disp_right_fname)
imgleft = np.array(result["left"])
np.save("./leftimg", imgleft)
disp_out = np.array(disp_left)
np.save("./displeft_out", disp_out)
disp_outr = np.array(disp_right)
np.save("./dispright_out", disp_outr)
occleft = np.array(occ_left)
np.save("./occleft", occleft)
occright = np.array(occ_right)
np.save("./occright", occright)
if self.split == "train":
# horizontal flip
#result['left'], result['right'], result['occ_mask'], result['occ_mask_right'], disp, disp_right \
# = horizontal_flip(result['left'], result['right'], occ_left, occ_right, disp_left, disp_right, self.split)
result['left'], result['right'], result['occ_mask'], result['occ_mask_right'], disp, disp_right \
=result['left'], result['right'], occ_left, occ_right, disp_left, disp_right
result['disp'] = np.nan_to_num(disp, nan=0.0)
result['disp_right'] = np.nan_to_num(disp_right, nan=0.0)
# random crop
result = random_crop(360, 640, result, self.split)
else:
result['occ_mask'] = occ_left
result['occ_mask_right'] = occ_right
result['disp'] = disp_left
result['disp_right'] = disp_right
result = augment(result, self.transformation)
#cc=result["disp"]
return result