yolov5-face

yolov5-face copied to clipboard

bbox is not correct after using rotation augmentation

1

I apply 60 degree image augmentation for training, "train_batch" images show that the bbox is wrong 

Iron-iii

As the title says, this model cannot currently be downloaded from Google Drive.

opsavi

Zero results on jetson platform

1

Hi, Jetson Xavier, cuda 11.4, TRT 8.4.1.5-1+cuda11.4. Followed the instructions, some warnings (see attached) but no error. The result.jpg has zero marking (no rectangles). [q1.txt](https://github.com/deepcam-cn/yolov5-face/files/10573732/q1.txt) Any Idea what is wrong...

danwanban

多角度训练问题，正面检测能力差了很多

2

The dataset was rotated by 3 angles for training, but the multi face detection ability at 0 degrees decreased, the face confidence decreased, and there were also false or missed...

chensheng-code

大佬tensorrt推理时main.py想使用jetson nano 的网络摄像头/dev/video0该如何改main.py的程序呀5555

zts2002

yolov5l-face.pt的google drive下载链接失效了，能否补一个链接

tenderness-git

请问忽略小脸该如何处理呢？

1

如题，需要在代码的哪些地方做一些改动？

sunjian2015

this model support TorchScript or not？I transform a TorchScript model, but the conf_thres value very low

WYZEYaolongZhu

有成功训练自己的数据集的大佬吗

9

由于项目要求，需要把眼睛上的关键点改为在眼角处标记，是不是只能重新对数据集进行标注？

1234dgfg

How does it work with images with different size than 640?

1

How does it work with images with different size than 640?

martinenkoEduard