songguanglu

Hi, we obtain the sensitive heatmap depends on the accuracy of the classification and regression of the region proposal in this location. Specifically, given a feature map H in the...

Hi, this is a very meaningful question. In fact, TSD does help alleviate the misalignment between classification and localization in RetinaNet. In our private experiments, we have improved the performance...

We have only performed the TSD variant on FCOS. If you are interested in the performance on RetinaNet, you can try to implement it based on this project. Any questions...

Hi，TSD project has not been adapted to MMdetV2.0 and the code is wrote under v1.x version (The old v1.x branch works with PyTorch 1.1 to 1.4). You can use v1.x...

Hi, we have the plan to share the pretrained weights on OpenImages dataset. It will adopt the SENet-154 backbone. We need to retrain this model on the mmdetection framework and...

> @songguanglu : did the training finish? It will help a lot of people, it kinda hard to find the pretrained Hi, we have finished the training for ResNet-50 with...

> > > @songguanglu : did the training finish? It will help a lot of people, it kinda hard to find the pretrained > > > > > > Hi,...

Hi, thanks for your advice and we will modify this in the future update.

Hi, the inference code of TSD is not adaptive to the demo/inference_demo.ipynb and we will modify it in the next commit. Can you train with the TSD configuration?

Hi, 'nvcc fatal : Unsupported gpu architecture 'compute_75'' seems that your cuda version not supports four your GPU. You can use a higher version of cuda such as cuda 10.x.