search

yz93 / LAVT-RIS

GNU General Public License v3.0
183 stars 14 forks source link

About Semantics-Aware Dynamic Localization #23

Closed PanXiongAdam closed 1 year ago

PanXiongAdam commented 1 year ago

Hello, Sir. I read your latest article(Semantics-Aware Dynamic Localization and Refinement for Referring Image Segmentation) which is not public code. I don't see you e-mail. So, I contact you in here. Can you tell me how to acheive the dynamic kernel K? It is very novel idea.

yz93 commented 1 year ago

Thanks for your interest in our work. It will be a while before we release the code.

Meanwhile, you can refer to https://github.com/PeizeSun/SparseR-CNN/blob/0d2909895fecca7051f0820328345fc61ff0a6e6/projects/SparseRCNN/sparsercnn/head.py#L258 for the implementation of the dynamic kernel.

jinx2018 commented 1 year ago

Thanks for your interest in our work. It will be a while before we release the code.

Meanwhile, you can refer to https://github.com/PeizeSun/SparseR-CNN/blob/0d2909895fecca7051f0820328345fc61ff0a6e6/projects/SparseRCNN/sparsercnn/head.py#L258 for the implementation of the dynamic kernel.

Hope you release the codes of paper Semantics-Aware Dynamic Localization and Refinement for Referring Image Segmentation soon

PanXiongAdam commented 9 months ago

Thanks for your interest in our work. It will be a while before we release the code.

Meanwhile, you can refer to https://github.com/PeizeSun/SparseR-CNN/blob/0d2909895fecca7051f0820328345fc61ff0a6e6/projects/SparseRCNN/sparsercnn/head.py#L258 for the implementation of the dynamic kernel.

@yz93 Hello, Sir. I re-implement the SADLR method, but it decreases the benchmarks of the results when I use the cross-entropy as the supervised loss and re-train the network. In this, we don't use the multi-class dice. So, can you tell me what caused this?