search

open-mmlab / mmdetection

OpenMMLab Detection Toolbox and Benchmark
https://mmdetection.readthedocs.io
Apache License 2.0
29.28k stars 9.41k forks source link

fsaf_x101_64x4d_fpn_1x_coco.py: /coco.py", line 230, in _det2json data['category_id'] = self.cat_ids[label] #3223

Closed ghost closed 4 years ago

ghost commented 4 years ago

Hello, This only happens at the end of the training loop and stops training after the first epoch: image

I tried both these solutions as suggested elsewhere:

pip install "git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI"
pip install -U "git+https://github.com/open-mmlab/cocoapi.git#subdirectory=pycocotools"

Thanks ,

ghost commented 4 years ago

This only happens if I try to validate while training (e.g. --validate)

HXH-resonance commented 4 years ago

This only happens if I try to validate while training (e.g. --validate)

请问您方便留一个联系方式吗？有些问题想向您请教。我的邮箱是Houxiaohan_666@163.com

ZwwWayne commented 4 years ago

This means that your results predict more labels and it is out of the range of the dataset's number of classes. This usually happens when you set the num_classes in the head greater than the real number of classes in the dataset.