MCG-NJU / MixFormer

[CVPR 2022 Oral & TPAMI 2024] MixFormer: End-to-End Tracking with Iterative Mixed Attention
https://arxiv.org/abs/2203.11082
MIT License
448 stars 75 forks source link

Model weights of MixViT-L #104

Open ymsun2020 opened 8 months ago

ymsun2020 commented 8 months ago

Hello, first of all thank you for the good work. Which model weight in the cloud drive corresponds to the MixViT-L model that achieves an AO of 75.7% on got10k? Is it mixformer_online_got_1k.pth.tar or mixformer_online_22k got.pth.tar?

yutaocui commented 8 months ago

Hi, we didn't upload the MixViT-Got model, but you can use MixViT-B for test.

Quan-w commented 6 months ago

Hi, we didn't upload the MixViT-Got model, but you can use MixViT-B for test.您好,安装环境还是和22年的mixformer一样嘛,我依照给予的install.sh安装timm==0.2.1,但是报错在mixfomer.py中from timm import MlP没有。

iZhuoxx commented 5 months ago

Hi, we didn't upload the MixViT-Got model, but you can use MixViT-B for test.

Hi, thank you for your sharing, could you please send the extraction password of Baidu cloud disk so we can get the model?

MaximeDebarbat commented 1 month ago

The password is shared on the README: hmuv