syscv / p3depth Goto Github PK

FileNotFoundError: [Errno 2] No such file or directory: '/dataset/nyu_depth_v2/official_splits/NYU_Depth_v2/data/nyu2_train.csv'

Hi, thanks for your great work. I'm just wondering when will you release the pre-trained models?
Bests

When I run the test, the mode loading shape does not match

Traceback (most recent call last):
File "trainer.py", line 228, in
main()
File "trainer.py", line 149, in main
model = DepthLitModel.load_from_checkpoint(CKPT_PATH, config=config)
File "/data/home/wuhaifeng/anaconda3/envs/p3depth/lib/python3.7/site-packages/pytorch_lightning/core/saving.py", line 157, in load_from_checkpoint
model = cls._load_model_state(checkpoint, strict=strict, **kwargs)
File "/data/home/wuhaifeng/anaconda3/envs/p3depth/lib/python3.7/site-packages/pytorch_lightning/core/saving.py", line 205, in _load_model_state
model.load_state_dict(checkpoint['state_dict'], strict=strict)
File "/data/home/wuhaifeng/anaconda3/envs/p3depth/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1672, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for DepthLitModel:
size mismatch for net.coords: copying a param with shape torch.Size([8, 480, 640, 2]) from checkpoint, the shape in current model is torch.Size([16, 480, 640, 2]).

In the function get_coords(), what is the meaning of the parameter fix_axis when it is set False?

Do you have any estimation on when the code and trained model are going to be released?

The results I get with kitti are all red, have you ever encountered this situation?

After setting up the environment, I used NYU dataset for training, but the training results were very strange. The loss function converged slowly, rmse kept increasing, and delta kept decreasing.

Hello,
Why do you divide by 255.0 for the NYU data values?
e.g. https://github.com/SysCV/P3Depth/blob/main/src/data/NYU_dataset.py#L82

The BTS loader does not do this https://github.com/cleinc/bts/blob/master/pytorch/bts_dataloader.py#L163C1-L166C52
Which make sense since the website says image depth values are already in meters. https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html