I ran into this problem when training with a custom data set.How to solve it?
Traceback (most recent call last):
File "tools/train.py", line 191, in
main()
File "tools/train.py", line 187, in main
meta=meta)
File "/content/drive/MyDrive/boxlevelset/mmdet/apis/train.py", line 172, in train_detector
runner.run(data_loaders, cfg.workflow)
File "/usr/local/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 127, in run
epoch_runner(data_loaders[i], **kwargs)
File "/usr/local/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 54, in train
self.call_hook('after_train_epoch')
File "/usr/local/lib/python3.7/site-packages/mmcv/runner/base_runner.py", line 307, in call_hook
getattr(hook, fn_name)(self)
File "/content/drive/MyDrive/boxlevelset/mmdet/core/evaluation/eval_hooks.py", line 147, in after_train_epoch
key_score = self.evaluate(runner, results)
File "/content/drive/MyDrive/boxlevelset/mmdet/core/evaluation/eval_hooks.py", line 177, in evaluate
results, logger=runner.logger, **self.eval_kwargs)
File "/content/drive/MyDrive/boxlevelset/mmdet/datasets/dcm.py", line 482, in evaluate
cocoEval.evaluate()
File "/usr/local/lib/python3.7/site-packages/pycocotools/cocoeval.py", line 149, in evaluate
self._prepare()
File "/usr/local/lib/python3.7/site-packages/pycocotools/cocoeval.py", line 110, in _prepare
_toMask(gts, self.cocoGt)
File "/usr/local/lib/python3.7/site-packages/pycocotools/cocoeval.py", line 95, in _toMask
rle = coco.annToRLE(ann)
File "/usr/local/lib/python3.7/site-packages/pycocotools/coco.py", line 497, in annToRLE
rles = maskUtils.frPyObjects(segm, h, w)
File "pycocotools/_mask.pyx", line 292, in pycocotools._mask.frPyObjects
IndexError: list index out of range
@ccdongxu Hello, the annotaion json file needs the coco format.
I upload the coco-style json file (Google driver) of VOC, you can refer to it.
@LiWentomng Thank you for your answer.The coco-style json file contains segmentation fields and bbox fields. Does this mean that both segmentation annotation and target box annotation are required in my json?
@ccdongxu The training process only needs the bbox fields in the train.json file. The val/test process need both segmentation
and box fields to evaluate the segmentatio performance (or bbox annotaion for detection performance, opitional.)
Do you mean that my val.json file must include segmentation fields?
But my dataset is marked by myself, labeled with labelme, and then converted to coco format. There are only four numbers in the segmentation, but the segmentation in the dataset used by the author has many numbers. How to solve this problem?
LiWentomng
