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
600 stars 108 forks source link

How to use the MS-AOT to test custom dataset #42

Open xwhkkk opened 1 year ago

xwhkkk commented 1 year ago

Hello !

Thanks for sharing your work ! How do use MS-AOT to evaluate custom tracking datasets and output the prediction mask ? Could you give me some suggestions ?

Thank you !

yoxu515 commented 1 year ago

Hi, please check here in README, which allows you to run models on custom datasets. The results are saved as videos and masks.

xwhkkk commented 1 year ago

Hi, please check here in README, which allows you to run models on custom datasets. The results are saved as videos and masks.

  • We haven't release the code of MS-AOT, but DeAOT might be a choice for better performance and efficiency compared with original AOT.

Thank you for your kind reply ! I would follow your suggestion. As my dataset has small targets and occlusions, I would also like to try evaluating it using the MS-AOT method. In addition, I noticed in issue #32 you mentioned the source code of MS-AOT, and I found in 'evaluate_ms_aot.sh' using vot-toolkit to get results. Can I use the checkpoint of MS-AOT and Mixformer to evaluate my own dataset following the 'tools/eval.py' ?

Thank you !

yoxu515 commented 1 year ago

Hi, MS-AOT currently is not integrated in this repository. The code in issue #32 is only for vot datasets and different from the framework in this repo. We will integrate the code of MS-AOT into this repo in recent days.

xwhkkk commented 1 year ago

Hi, MS-AOT currently is not integrated in this repository. The code in issue #32 is only for vot datasets and different from the framework in this repo. We will integrate the code of MS-AOT into this repo in recent days.

Thank you for your kind reply !

asaini68-gtri commented 11 months ago

Any update on MS-AOT code availability?