I have trouble with these bugs

[lindermanfire]

Thank you for your dedication and generosity in making this work available. It has been incredibly insightful and beneficial for my own projects.

when I use python train.py --config configs/default.txt --data_dir D:\OmniTrackFast\dataset\lab-coat there are some bugs:
raft_online.py

……
for i in tqdm.trange(self.num_imgs):
for j in range(-16, 17):
if j == 0:
continue
if i + j < 0 or i + j >= self.num_imgs:
continue
flow_file = os.path.join(self.flow_dir, '{}_{}.npy'.format(self.img_names[i], self.img_names[i+j]))
flow = np.load(flow_file)
mask_file = flow_file.replace('raft_exhaustive', 'full_mask').replace('.npy', '.png')
masks = imageio.imread(mask_file) / 255.
if masks.sum() < 4096:
continue
self.candidate_pair_range[i, 0] = min(self.candidate_pair_range[i, 0], j)
self.candidate_pair_range[i, 1] = max(self.candidate_pair_range[i, 1], j)
if j > 0:
self.flow[i, j+15] = flow
self.masks[i, j+15] = masks

            else:
                self.flow[i, j+16] = flow
                **self.masks[i, j+16] = masks**

……
self.masks[i, j+16] shape is 2D,masks shape is 3D,is it change to self.masks[i, j+16] = masks[:,:,0]?

Then in trainer_combo.py it assert:
……
depth_files = depth_files[:self.num_imgs]
assert len(depth_files) == self.num_imgs
……
So I comment it to
……
depth_files = depth_files[:self.num_imgs]
#assert len(depth_files) == self.num_imgs
……
but then assert it:
assert x_range > 0 and y_range > 0 and z_range > 0
I dont know how to solve it,can you help me to solve it.

[lindermanfire]

感谢您的工作，您的工作对我的启发非常大，但可能由于我对部署方法不熟悉，现在碰见了以下一些问题:
当我使用python train.py --config configs/default.txt --data_dir D:\OmniTrackFast\dataset\lab-coat (lab-coat官方数据集)训练时 碰见以下报错:
……
for i in tqdm.trange(self.num_imgs):
for j in range(-16, 17):
if j == 0:
continue
if i + j < 0 or i + j >= self.num_imgs:
continue
flow_file = os.path.join(self.flow_dir, '{}_{}.npy'.format(self.img_names[i], self.img_names[i+j]))
flow = np.load(flow_file)
mask_file = flow_file.replace('raft_exhaustive', 'full_mask').replace('.npy', '.png')
masks = imageio.imread(mask_file) / 255.
if masks.sum() < 4096:
continue
self.candidate_pair_range[i, 0] = min(self.candidate_pair_range[i, 0], j)
self.candidate_pair_range[i, 1] = max(self.candidate_pair_range[i, 1], j)
if j > 0:
self.flow[i, j+15] = flow
self.masks[i, j+15] = masks
else:
self.flow[i, j+16] = flow
self.masks[i, j+16] = masks

……
self.masks[i, j+16] shape 是二维的,但 masks shape 是三维的,是否该改成 self.masks[i, j+16] = masks[:,:,0]?

接着后边出现报错:
trainer_combo.py：
……
depth_files = depth_files[:self.num_imgs]
assert len(depth_files) == self.num_imgs
……
所以我把第二行注释了
……
depth_files = depth_files[:self.num_imgs]
#assert len(depth_files) == self.num_imgs
……
但之后仍出现报错
assert x_range > 0 and y_range > 0 and z_range > 0
而后我不知道该如何去修改这个BUG，希望您在百忙之中能否抽空帮我看下这个问题.

[TimSong412]

Hi!
Do you have valid depth files stored in the correct path? It seems to be some problems with loading depths.

Sorry for the late reply, thanks very much for you interest!

[lindermanfire]

Hi! Do you have valid depth files stored in the correct path? It seems to be some problems with loading depths.

Sorry for the late reply, thanks very much for you interest!

Thank you so much! So how can I get depth files with dataset lab-coat?

[TimSong412]

Hi! Could you refer to the file preprocessing/README.md
There is instructions on how to get the flow and depth

[lindermanfire]

Hi! Could you refer to the file preprocessing/README.md There is instructions on how to get the flow and depth

Fllow by the instructions preprocessing/README.md then I cant find the file "download_models.sh" ,and when I use python main_processing.py --data_dir --chain ,it report errors where "from raft import RAFT" and "from utils.utils import InputPadder"