How different is this paper

[jaideep11061982]

https://arxiv.org/pdf/2211.08217v1.pdf

@zhiyuanyou could you read through this once. author has explained the logic . for attention feature they are using I think Transformer encoder layer . Later part not clear to me. So wanted to understand if there is big or minor difference between your paper and this new one

[zhiyuanyou]

Well, I had read this paper in the past. I think this paper proposes a better method considering its outstanding performance.

[jaideep11061982]

@zhiyuanyou any small help I can get in implementing its Object protoptype extraction part, It appears analogous to your Safe count block logic After reading through this paper , it appears they use Resnet 50 as backbone followed by Multiheaded attention module which then inputs to Object prototype extraction module (this part is where I face some challenges ) , it appears it tryign to do similar thing as you do in SafeCountMultiblock layer of model

[zhiyuanyou]

It is better to contact the authors for their code.