VISION-SJTU / USOT

[ICCV2021] Learning to Track Objects from Unlabeled Videos
63 stars 7 forks source link

怎么获得预训练模型? #10

Closed chenrxi closed 2 years ago

chenrxi commented 2 years ago

你好,请问预训练模型是在什么数据上训练得到的吗?有预训练的超参吗?

zhengjilai commented 2 years ago

Q: how the pretrained model is obtained ?

A: For the supervised backbone (imagenet_pretrain), it is directly collected from TracKit as Ocean is our base tracker. For the self-supervised backbone (moco v2), it is directly collected from the MoCo project.

chenrxi commented 2 years ago

thank you for your answer