Hi: when we use your tool img_split.py, meet the error TypeError: can't pickle _thread.RLock objects, happens " patch_infos = pool.map(worker, zip(infos, img_dirs))" , looks like multiprocessing problem , can you suggest us how to solve the problem， thanks

When use img_split.py with my own dataset, I get error as below. After add Image.MAX_IMAGE_PIXELS = None to DOTAio.py, the error has gone. Do you think is there any cons when adding this line?

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "BboxToolkit/tools/img_split.py", line 316, in <module>
    main()
  File "BboxToolkit/tools/img_split.py", line 268, in main
    nproc=args.nproc)
  File "/obbdetection/BboxToolkit/BboxToolkit/datasets/DOTAio.py", line 34, in load_dota
    contents = pool.map(_load_func, os.listdir(img_dir))
  File "/opt/conda/lib/python3.7/multiprocessing/pool.py", line 268, in map
    return self._map_async(func, iterable, mapstar, chunksize).get()
  File "/opt/conda/lib/python3.7/multiprocessing/pool.py", line 657, in get
    raise self._value
PIL.Image.DecompressionBombError: Image size (625847040 pixels) exceeds limit of 178956970 pixels, could be decompression bomb DOS attack.

I want to generate annotation .txt for splitted image. but, img_split.py create just .pkl files.

So, I create new file img_split_dota.py file. It can make the annotation .txt file for spllitted image.

Please review this PR.

Considering your feedback, I try to merging the function from img_split_dota.py to img_split.py.

To use classname in img_split.py, I did modify the BboxToolkit/datasets/DOTAio.py

please review.

This is my configuration file when I visualized Dior-R , { "load_type": "pkl", "img_dir": "/data/dataset/DIOR/JPEGImages-test", "ann_dir": "/data1/bk/6742.pkl", "classes": "dior", "prior_annfile": null, "merge_type": "addition", "load_nproc": 10,

"skip_empty": false,
"random_vis": false,
"ids": null,
"show_off": true,
"save_dir": "vis_bk/dior",
"vis_nproc": 10,

"shown_btype": null,
"shown_names": "./vis_configs/dior/short_names.txt",
"score_thr": 0.2,
"colors": "./vis_configs/dior/colors.txt",
"thickness": 3.0,
"text_off": true,
"font_size": 10,
"wait_time": 0

} The following error occurs: Traceback (most recent call last): File "visualize.py", line 230, in main() File "visualize.py", line 143, in main nproc=args.load_nproc) File "/data/bk/OBBDetection/BboxToolkit/BboxToolkit/datasets/io.py", line 54, in load_pkl old_classes, contents = data['cls'], data['content'] TypeError: list indices must be integers or slices, not str

I output the data parsed by .pkl this is

array([[ 7.8454449e+02, 6.3927423e+02, 2.6410254e+01, 8.1578226e+00, -1.2265325e-02, 7.9247594e-01], [ 5.6819104e+02, 6.5477740e+02, 1.5486428e+01, 8.1747618e+00, 1.5659013e+00, 7.7006650e-01], [ 5.7608038e+02, 6.5416187e+02, 1.5025672e+01, 8.0598936e+00, 1.5541477e+00, 6.8269205e-01], [ 5.7347760e+02, 6.9571722e+02, 1.3365779e+01, 5.5745769e+00, -1.5615332e+00, 4.7462189e-01], [ 5.7240784e+02, 6.5443042e+02, 1.5609086e+01, 1.1678979e+01, 1.5074093e+00, 4.4766316e-01], [ 5.3823401e+02, 6.5405450e+02, 1.7242418e+01, 1.0737660e+01, -1.5569979e+00, 2.6717269e-01], [ 6.1477234e+02, 1.3902811e+02, 3.6538612e+01, 8.9332752e+00, 1.4462376e-01, 1.9412568e-01], [ 5.7195233e+02, 6.5014233e+02, 1.4650148e+01, 8.8534908e+00, 5.1303148e-02, 1.0566589e-01], [ 5.4264069e+02, 6.5394983e+02, 1.8320833e+01, 1.6436756e+01, 1.5667925e+00, 5.3098477e-02]], dtype=float32), array([], shape=(0, 6), dtype=float32)] I would like to ask you what is the problem, looking forward to your reply

If I don't get it wrong, if the iou between hbbox of the objects and windows is less than the value of iou_thr, the object will be ignored. Right? I wonder why you do this?