search

MIV-XJTU / ARTrack

Apache License 2.0
202 stars 32 forks source link

Altered decoder #69

Open iMaTzzz opened 1 month ago

iMaTzzz commented 1 month ago

Hello ! I was looking through your paper and your code again and I was wondering where is the actual altered decoder ? Is it already implemented by default or do I need to do some modifications in the layers ? Furthermore, are these changes compatible with the current checkpoints you are providing ? Thanks in advance

AlexDotHam commented 1 month ago

The alter decoder's pipeline does not provide, actually, you only need to change the decoder and seq_decoder's mask attention and pipeline in a simple way. the checkpoints we provide are suitable for traditional decoders, we consider to provide a alter version in future maybe one month later, because I need to finish my paper in NeurIPS 2024.

iMaTzzz commented 1 month ago

Thanks for the answer. That would be actually great if that's possible, thank you for taking time for this. Meanwhile. could you also let this issue open so that when you want to add the altered version. I would be more than happy to receive an update.

AlexDotHam commented 1 month ago

That will be fine

iMaTzzz commented 4 days ago

Hello, I hope the CVPR 2024 went well for you ! I wanted to know if there have been any updates on the Altered Decoder and ARTrackV2. Looking forward to any news you can share. Thanks !

AlexDotHam commented 2 days ago

I really don't have time to release the alter decoder code, so sorry for that, I think it is easier to reproduce that if you are familiar with this code. But the ARTrackV2's code will be released when I finish the ARTrackV3, this will be soon, you will see that in CVPR2025, and the code of ARTrackV2 will be released as a V3 version which is not the same to V2(prevent from the patent legal risks).

iMaTzzz commented 2 days ago

No problem I understand, thank you for taking the time to answer as well as sharing some part of your work !

Do you have an approximate time on the release date on the V3 version ? I don't want to implement the altered decoder if you are going to release a state of the art model.

Thanks again !