How to create or acquire training/label_3 directory?

[maheriya]

I am trying to run eval using

sh scripts/eval_with_trained_model.sh

I have downloaded the pre-trained pob model, and moved to data/kitti/pob as per instruction. However, the eval script fails as it doesn't find 'training/label_3' directory. Where can I find this directory? KITTI dataset doesn't have this. Are you generating this somehow?

Here is the excerpt of the output with errors:

...
...
TEST:
  DETECTIONS_PER_IMG: 100
  EXPECTED_RESULTS: []
  EXPECTED_RESULTS_SIGMA_TOL: 4
  FORCE_RECOMPUTE: True
  IMS_PER_BATCH: 4
2020-05-29 16:05:50,813 disprcnn.utils.checkpoint INFO: Loading checkpoint from models/kitti/pob/smrcnn.pth
2020-05-29 16:05:51,135 disprcnn.data.build WARNING: When using more than one image per GPU you may encounter an out-of-memory (OOM) error if your GPU does not have sufficient memory. If this happens, you can reduce SOLVER.IMS_PER_BATCH (for training) or TEST.IMS_PER_BATCH (for inference). For training, you must also adjust the learning rate and schedule length according to the linear scaling rule. See for example: https://github.com/facebookresearch/Detectron/blob/master/configs/getting_started/tutorial_1gpu_e2e_faster_rcnn_R-50-FPN.yaml#L14
generating /IMAGESETS/work/DispRCNN/data/kitti/object/training/label_3/annotations.pkl
  0%|                                                                                                                      | 0/7481 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "tools/test_net.py", line 97, in <module>
    main()
  File "tools/test_net.py", line 80, in main
    data_loaders_val = make_data_loader(cfg, is_train=False, is_distributed=distributed)
  File "/IMAGESETS/work/DispRCNN/disprcnn/data/build.py", line 147, in make_data_loader
    cfg.SOLVER.OFFLINE_2D_PREDICTIONS)
  File "/IMAGESETS/work/DispRCNN/disprcnn/data/build.py", line 45, in build_dataset
    dataset = factory(**args)
  File "/IMAGESETS/work/DispRCNN/disprcnn/data/datasets/kitti.py", line 323, in __init__
    super().__init__(root, split, transforms, filter_empty, offline_2d_predictions_path, 'pob')
  File "/IMAGESETS/work/DispRCNN/disprcnn/data/datasets/kitti.py", line 47, in __init__
    self.annotations = self.read_annotations()
  File "/IMAGESETS/work/DispRCNN/disprcnn/data/datasets/kitti.py", line 173, in read_annotations
    anno_per_img = load_label_3(self.root, 'training', i)
  File "/IMAGESETS/work/DispRCNN/disprcnn/utils/kitti_utils.py", line 262, in load_label_3
    with open(absolute_path) as f:
FileNotFoundError: [Errno 2] No such file or directory: '/IMAGESETS/work/DispRCNN/data/kitti/object/training/label_3/000000.txt'
Traceback (most recent call last):
  File "tools/split_predictions.py", line 36, in <module>
    main()
  File "tools/split_predictions.py", line 32, in main
    split(dir_name, split_path)
  File "tools/split_predictions.py", line 11, in split
    predictions = torch.load(os.path.join(dir_name, 'predictions.pth'), 'cpu')
  File "/IMAGESETS/work/anaconda/envs/disprcnn/lib/python3.7/site-packages/torch/serialization.py", line 381, in load
    f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: 'models/kitti/pob/inference/kitti_val_pob/predictions.pth'
2020-05-29 16:05:53,418 disprcnn INFO: Using 1 GPUs
...
...

I don't have 'data/kitti/object/training/label_3' directory. Could you help?

[ootts]

Duplicate issue. See here for detail.

[maheriya]

Thanks. I should have searched harder before opening the issue!