Haoyi Zhu

icon indicating a website link https://www.haoyizhu.site/

icon company Shanghai Artificial Intelligence Laboratory icon location Shanghai, China icon location 2nd-year Ph.D. student @ Shanghai AI Lab & USTC, advised by Prof. Tong He, Prof. Wanli Ouyang & Prof. Xiaogang Wang. World Model & Embodied AI.

Results 41 comments of Haoyi Zhu

single hand output error

Singlehand版本的模型需要提供单手的大致bbox，而不能直接用人体的bbox。见Model ZOO当中 `If you want to use the single hand model, you should give the rough bounding box of a single hand instead of that of a whole person.`

single hand output error

如果你对单手bbox的获取有困难的话，我建议可以直接用全身的模型检测手

1%|█▏ | 21/1423 [00:02<02:09, 10.81it/s]已杀死

可能是内存不够 linux在内存不够的情况下可能会自动把进程kill掉

about FastPose Paper

Thanks! Our paper is coming soon!

How to install only CPU on Linux

Hi. You could try to run `scripts/demo_inference.py` with flag `--gpus -1` which means for cpu only.

use detfile

Hi, you can use the `--vis` or `--save_img` flag when running `scripts/demo_inference.py`.

Cancelling YOLO and feed the pose estimator via another source

Hi, you can try the [detfile](https://github.com/MVIG-SJTU/AlphaPose/blob/master/scripts/demo_inference.py#L36) flag and the [file detection loader](https://github.com/MVIG-SJTU/AlphaPose/blob/master/alphapose/utils/file_detector.py#L15), so that you can just send into your detection result file.

Change backbone to mobilenetv3

I'm not familiar with mobilenet, but it seems the dim of input to pixel_shuffle is wrong. Could you print it's shape out?

Training Process Stuck

I'm not sure about the reason. Could you try to run `python ./scripts/train.py --exp-id ${EXPID} --cfg ${CONFIG} --sync`? Or could you try to reduce the thread number by `--nThreads 0`?...

will you support 3D pose?

Hi, 3D pose is supported now!