exely / masktrackrcnn-mmdet2.0 Goto Github PK

I want to know how to do the first step mentioned in training part about dataset path. What should be done here
urgent!!!

When I finish configuring the environment and start training according to the requirements in the document, the following errors are reported:
ImportError: /home/zhangsai/.conda/envs/MaskTrackRCNN/lib/python3.8/site-packages/mmcv/_ext.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZN3c105ErrorC1ENS_14SourceLocationENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE
My environment configuration is as follows:
python:3.8.12
pytorch:1.7.1
torchvision:0.8.2
cudatoolkit:11.0.221
mmcv-full:1.3.14
mmdet:2.11.0
ubuntu:18.04
I want to know what went wrong. I look forward to your answer. Thank you very much!!!

When I run commands

python tools/train.py configs/masktrackrcnn/masktrackrcnn_r50_fpn_2x.py

After training for an epoch, it raised exception:
NotImplementedError: evaluation not exists!

Then I found that evaluate function in ytvos.py file was not implemented yet.
I want to how to evaluate the validation set, or how the evaluate function could be implemented?

Hi. I was trying out the test phase. My model was the same model produced in Project MaskTrackRCNN. that's mean epoch_12.pth The test phase is done well. But when running the code, python scripts/json2mask_v2.py
the outputs are as follows. What's wrong with that?

sample output:
https://drive.google.com/file/d/1cIzu-i2RKr2TQ-yj6dlCSk4j4CD2kxn1/view?usp=sharing
https://drive.google.com/file/d/1FNEhVT6cqrGC0MeMkkqq7egSjHTdJNEl/view?usp=sharing

请问这个代码使用的数据集是哪个？

可以从哪里下载？

感谢！

Why the num_classes is set to be 1 instead of 41 or 40?

I just want to test this algorithms on my own dataset, could you please release the pretrained model?Or share the pretrained model with me?Thanks!