search

Event-AHU / OpenPAR

[OpenPAR] An open-source framework for Pedestrian Attribute Recognition, based on PyTorch
MIT License
85 stars 11 forks source link

Accuracy and F1 score when using ViT-base #5

Closed MINDYEOI closed 10 months ago

MINDYEOI commented 10 months ago

Hi. Thank you for sharing your great work.

I wonder how much is the mA and F1 score if SequencePAR and PromptPAR backbone is changed into ViT-b/16 on PA100k, PETA, RAPv2 dataset. If you conducted the experiments, could you kindly share in public?

Thank you.

1125178969 commented 10 months ago

Sorry, we have not experimented with replacing the backbone on more datasets, if you are interested, you can try to reproduce the code for your experiments, or we will update these results later!

MINDYEOI commented 10 months ago

Thank you :)