search

yoxu515 / aot-benchmark

An efficient modular implementation of Associating Objects with Transformers for Video Object Segmentation in PyTorch
BSD 3-Clause "New" or "Revised" License
592 stars 107 forks source link

The modules that work on tiny objects #50

Open xwhkkk opened 1 year ago

xwhkkk commented 1 year ago

Hello !

Thanks for sharing your great work!

In section 6.1, the paper mentioned the R50-DeAOT-L performs better than R50-AOT-L on tiny or scale-changing objects. I would like to know which module is beneficial to tiny or scale-changing objects.

Looking forward to your kind reply !

z-x-yang commented 1 year ago

The performance gains mainly come from the decoupling of visual and ID embeddings. In fact, the improvement is general and not limited to tiny or scale-changing objects. I highlighted those examples since R50-AOT-L usually fails in those cases.