mmdetection: training of maskrcnn not work at all.
Describe the issue
I try to use my own dataset which has the coco format, and use them for training a mask Rcnn model. I followed the step in the Tutorial. I modify the config files, and have my new configuration in a path (configs/human).
Reproduction
-
What command or script did you run? i do the same as this tutorial (https://mmdetection.readthedocs.io/en/latest/2_new_data_model.html) i used the training command ‘’!python tools/train.py configs/human/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_person.py’’ and i have only one class.
-
What config dir you run? i used a new director contains all three main config needs in maskRcnn detector;
- Did you make any modifications on the code or config? Did you understand what you have modified?
- What dataset did you use? my own small sample of my dataset in coco form.
Environment
- Please run
python mmdet/utils/collect_env.pyto collect necessary environment information and paste it here. i did it in colab
could it be a problem in mmcv installation? how can i be sure it is like that.
About this issue
- Original URL
- State: closed
- Created 3 years ago
- Comments: 21 (5 by maintainers)
The dataset_type should be your dataset class name. Are you creating a new dataset class? If not, there is no need to change the dataset_type. I still think the problem is about the JSON file.
I have met the same problem,finally I figure out it is caused by the dataset config in the configs. The dataset_type should be set as ‘mmdet.CocoDatasets’
TypeError: init() missing 2 required positional arguments: 'doc' and 'pos'means that your JSON file is still incorrect. Refer to https://docs.python.org/3/library/json.html#json.JSONDecodeError I’m not sure what caused this, but I found that there is(1)in your json filename. Try to rename it.