boostcampaitech3 / level2-data-annotation_cv-level2-cv-09

level2-data-annotation_cv-level2-cv-09 created by GitHub Classroom
0 stars 4 forks source link

[Data] Annotation 실습 결과물 학습(poly_ill) #37

Open updaun opened 2 years ago

updaun commented 2 years ago

What

Why?

Todo

작업 내용

  1. git develop에서 최신 버전을 pull
  2. poly_ill.json 서버에 넣고 이름을 train.json으로 변경
  3. dataset.py 를 example/polygon/dataset.py로 변경

실험 파라미터 : 기본

데이터셋 디렉토리 구조

input/
`-- data
    |-- ICDAR17_Korean(기본 제공 데이터)
    |   |-- images
    |   `-- ufo
    |       `-- train.json
    |-- dataset(추가 annotation 데이터)
    |   |-- images
    |   `-- ufo
    |       |-- train.json
    |       `-- train_origin.json
    `-- dataset.zip

7 directories, 1828 files

Result

Data f1 recall precision
ICDAR17_Korean 0.4843 0.3716 0.6949
실습 Annotation Poly to Rect 0.4864 0.4037 0.6117
실습 Annotation Poly to Rect revised 0.5803 0.4899 0.7118
실습 Annotation Poly 포함 이미지 제거 0.4420 0.3548 0.5459
실습 Annotation Poly illegibility True 0.4734 0.3658 0.6708
jeongjae96 commented 2 years ago

image

다운님 ICDAR 데이터가 학습된 것으로 보입니다

updaun commented 2 years ago

저 학습 잘못 시켰네용.. config 파일 저장 안하고 돌려버렸네용.. 다시 학습 시작합니다..

updaun commented 2 years ago
[Epoch 1]:  67%|████████████████████████████▉              | 72/107 [05:11<02:31,  4.33s/it, Cls loss=0.824, Angle loss=0.165, IoU loss=1.59, _timestamp=1.65e+9, _runtime=330]
Traceback (most recent call last):
  File "train.py", line 149, in <module>
    main(args)
  File "train.py", line 142, in main
    do_training(**args.__dict__)
  File "train.py", line 87, in do_training
    for img, gt_score_map, gt_geo_map, roi_mask in train_loader:
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 435, in __next__
    data = self._next_data()
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1065, in _next_data
    return self._process_data(data)
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1111, in _process_data
    data.reraise()
  File "/opt/conda/lib/python3.8/site-packages/torch/_utils.py", line 428, in reraise
    raise self.exc_type(msg)
FileNotFoundError: Caught FileNotFoundError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 198, in _worker_loop
    data = fetcher.fetch(index)
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/opt/ml/code/east_dataset.py", line 136, in __getitem__
    image, word_bboxes, roi_mask = self.dataset[idx]
  File "/opt/ml/code/dataset.py", line 379, in __getitem__
    image = Image.open(image_fpath)
  File "/opt/conda/lib/python3.8/site-packages/PIL/Image.py", line 2975, in open
    fp = builtins.open(filename, "rb")
FileNotFoundError: [Errno 2] No such file or directory: '../input/data/dataset/images/17_11_26 오전 9_55 Office Lens.jpg'

파일 이름에 한글이 있으면 디렉토리를 못 찾아오는 것 같습니다!

updaun commented 2 years ago

이미지 파일에 있는 해당 사진의 이름을 17_11_26 #Uc624#Uc804 9_55 Office Lens.jpg -> 17_11_26 오전 9_55 Office Lens.jpg 변경해서 학습 진행하니 오류나지 않고 학습 진행 하고 있습니다.