Issues of syntax, API change, code errors or exceptions will be addressed immediately.
If you find it difficult to reproduce an experiment, please run evaluation of the released models first.
Simplicity comes at a cost of performance. For a faster version, checkout version v5.1.
We don't want a repository as historic archive only. However, we cannot run experiment after every change. To the best of our knowledge, both v5.1 and v6 can reproduce the results above.
For training on your own dataset, please check intermediate steps carefully. Out of the box training and inference may not apply.
We want a repo that supports both Gluon and Symbolic training. dev branch is our attempt and it is very close. You are absolutely welcome to continue this effort. For now, our best effort is https://gluon-cv.mxnet.io/model_zoo/index.html.
dev
branch is our attempt and it is very close. You are absolutely welcome to continue this effort. For now, our best effort is https://gluon-cv.mxnet.io/model_zoo/index.html.