Please help me with use vid4 and vimo90k

[luoyinTT]

(1)What is the file structure of vimo90k,
(2)when i use vid4 for basicvsr to test , i always meet :

Original Traceback (most recent call last):

  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 202, in _worker_loop
    data = fetcher.fetch(index)
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/base_sr_dataset.py", line 52, in __getitem__
    return self.pipeline(results)
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/pipelines/compose.py", line 43, in __call__
    data = t(data)
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/pipelines/loading.py", line 135, in __call__
    f'filepath should be list, but got {type(filepaths)}')
TypeError: filepath should be list, but got <class 'str'>

in the config file:

test_dataset_type = 'SRVid4Dataset'
# test
    test=dict(
        type=test_dataset_type,
        lq_folder='videoSR/BIx4',
        gt_folder='videoSR/original',
        ann_file='videoSR/annotation.txt',
        pipeline=test_pipeline,
        scale=4,
        filename_tmpl='{:03d}',
        num_input_frames=7,
        test_mode=True),
)

[ckkelvinchan]

1. Vimeo-90K has a structure of clip/sequence/img.png. For instance,

| 00001
    | 0001
        | im1.png
        | im2.png
             ...
    | 0002
      ...
| 00002
   ...

2. SRVid4Dataset is for sliding-window methods such as EDVR. Please use SRTestMultipleGTDataset. For more details, please refer to the validation settings in basicvsr_vimeo90k_bd.py.

[innerlee]

@ckkelvinchan please add folder structure in the doc https://mmediting.readthedocs.io/en/latest/sr_datasets.html#vimeo90k-dataset

[luoyinTT]

@ckkelvinchan
i change config file:

test_dataset_type = 'SRTestMultipleGTDataset'

test_pipeline = [
    dict(
        type='LoadImageFromFileList',
        io_backend='disk',
        key='lq',
        channel_order='rgb'),
    dict(
        type='LoadImageFromFileList',
        io_backend='disk',
        key='gt',
        channel_order='rgb'),
    dict(type='RescaleToZeroOne', keys=['lq', 'gt']),
    dict(type='MirrorSequence', keys=['lq']),
    dict(type='FramesToTensor', keys=['lq', 'gt']),
    dict(
        type='Collect',
        keys=['lq', 'gt'],
        meta_keys=['lq_path', 'gt_path', 'key'])
]

test=dict(
        type=test_dataset_type,
        lq_folder='videoSR/BIx4',
        gt_folder='videoSR/original',
        pipeline=test_pipeline,
        scale=4,
        test_mode=True),
)
the same as SRTestMultipleGTDataset

- my vid4 dataset structure is 

| videoSR 
     | BIx4
          | city
             | Frame001.jpg
             ...
          ...
     | original
          | city
             | Frame001.jpg
             ...        
          ...

but i still meet the problem:

Traceback (most recent call last):
  File "tools/test.py", line 136, in <module>
    main()
  File "tools/test.py", line 99, in main
    save_image=args.save_image)
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/apis/test.py", line 40, in single_gpu_test
    for data in data_loader:
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 517, in __next__
    data = self._next_data()
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1199, in _next_data
    return self._process_data(data)
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1225, in _process_data
    data.reraise()
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/_utils.py", line 429, in reraise
    raise self.exc_type(msg)
TypeError: Caught TypeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 202, in _worker_loop
    data = fetcher.fetch(index)
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/scratch/workspace/yinl/anaconda3/envs/open-mmlab/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/base_sr_dataset.py", line 52, in __getitem__
    return self.pipeline(results)
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/pipelines/compose.py", line 43, in __call__
    data = t(data)
  File "/group/dphi_algo_scratch_03/yinl/mmediting-master/mmedit/datasets/pipelines/loading.py", line 135, in __call__
    f'filepath should be list, but got {type(filepaths)}')
TypeError: filepath should be list, but got <class 'str'>

[ckkelvinchan]

Your test_pipline is not correct. You should use

test_pipeline = [
    dict(type='GenerateSegmentIndices', interval_list=[1], filename_tmpl='Frame{:03d}.jpg', start_index=1),
    dict(
        type='LoadImageFromFileList',
        io_backend='disk',
        key='lq',
        channel_order='rgb'),
    dict(
        type='LoadImageFromFileList',
        io_backend='disk',
        key='gt',
        channel_order='rgb'),
    dict(type='RescaleToZeroOne', keys=['lq', 'gt']),
    dict(type='FramesToTensor', keys=['lq', 'gt']),
    dict(
        type='Collect',
        keys=['lq', 'gt'],
        meta_keys=['lq_path', 'gt_path', 'key'])
]

Moreover, please note the use of filename_tmpl='Frame{:03d}.jpg' and start_index=1 in GenerateSegmentIndices. Please find more details here.

[ckkelvinchan]

@ckkelvinchan please add folder structure in the doc https://mmediting.readthedocs.io/en/latest/sr_datasets.html#vimeo90k-dataset

Okay.

[luoyinTT]

@ckkelvinchan
thank you for your reply, i also want to know if i use vimo90k, do i have to run the preprocess_vimeo90k_dataset.py ?

and if i use vimeo90k to train basicvsr
the train_pipline and data train is right like this ?

mmedit/models/restorers/basic_restorer.py

[ckkelvinchan]

Yes, you can use preprocess_vimeo90k_dataset.py to generate the annotation file.

And you can follow exactly basicvsr_vimeo90k_bi.py for training on Vimeo-90K.

[luoyinTT]

@ckkelvinchan thank you ^^

[ckkelvinchan]

Closing due to inactivity, please reopen if there are any further problems.