search

NVlabs / neuralrgbd

Neural RGB→D Sensing: Per-pixel depth and its uncertainty estimation from a monocular RGB video
Other
301 stars 41 forks source link

Error in training on kitti dataset - valid batches check fail #4

Closed Ahmadgafer closed 5 years ago

Ahmadgafer commented 5 years ago

Hi, I'm trying to run the training on the kitti dataset. I downloaded the raw data from the official website. The testing is working for me but when I launch the training script I get: Traceback (most recent call last): File "train_KVNet.py", line 384, in <module> main() File "train_KVNet.py", line 323, in main loss_v = LOSS[-1] IndexError: list index out of range I tried to trace the problem and found that the condition in line 288 in train_KVnet.py always doesn't hold. if n_valid_batch > 0: I don't think I did something wrong as the testing is working on the same directory. Any ideas on how to solve this issue?

cxlcl commented 5 years ago

This may indicate that the loader didn't find the raw images. Could you set a breakpoint before if n_valid_batch > 0:, and look at the values for local_info['ref_dats'][0]['scene_path'] and local_info['ref_dats'][0]['img_path'] ?

They should be something like /datasets/kitti/rawdata/2011_09_26 and '/datasets/kitti/rawdata/2011_09_26/2011_09_26_drive_0001_sync/image_02/data/0000000007.png'

Ahmadgafer commented 5 years ago

@cxlcl Thank you for your reply. I will check it.

777-en commented 2 years ago

This may indicate that the loader didn't find the raw images. Could you set a breakpoint before if n_valid_batch > 0:, and look at the values for local_info['ref_dats'][0]['scene_path'] and local_info['ref_dats'][0]['img_path'] ?

They should be something like /datasets/kitti/rawdata/2011_09_26 and '/datasets/kitti/rawdata/2011_09_26/2011_09_26_drive_0001_sync/image_02/data/0000000007.png'

Hello, I met the same problem as @Ahmadgafer and I checked local_info['ref_dats'][0]['scene_path'] and local_info['ref_dats'][0]['img_path'], which are all right. Is there any other possible reasons for this issue? please.