zt1112
Results
3
comments of
zt1112
> > Thank you. We use [vid2vid](https://github.com/NVIDIA/vid2vid) to perform translation from skeleton to realistic videos. You need to prepare a small collection of motion/dancing videos as well as extracted skeleton...
@CaoWGG 你好，我设置的是0.01，因为设置0.0会报错，同样对比的是原版设置thresh0.01的结果
@CaoWGG 因为用在行人检测上，所以只测了cocoval的human类别。原版的是52.5mAP，tensorRT fp32是51.6mAP，tensorRT int8是50.5mAp