ncnn_Android_face

ncnn_Android_face copied to clipboard

I mean this model:https://github.com/FeiGeChuanShu/ncnn_Android_face/blob/main/ncnn-android-scrfd-master/app/src/main/assets/facemesh-op2.param

wwdok

cheng-orange

你好，FaceMesh输出的关键点位置不对 如果把脸位置超出左上角，看到模型输出的坐标没有负数 在MidiaPipe中是有负数的 我的手机是小米11 

liuguicen

Hello @FeiGeChuanShu, can be change input shape of FaceSegment model?

make-stronger

Hello, Great work with this! We're a team of Computer vision developers and work with TNN, Ncnn + MNN - and I'd like to have your email for some further...

ofirkris

I have downloaded the repo and set up the opencv and ncnn. opencv: SET(OpenCV_VERSION 4.5.1) nccn version: ncnn-20210322-android-vulkan NDK Version: 22.1.7171670 The build compilation is successful but when I try...

khkred

Could NOT find Threads (missing: Threads_FOUND) Call Stack (most recent call first):

Tanlang180

请教一下您是怎么把pth模型转换为ncnn的，我转完模型后出现了很大的偏差

1

Iron-iii

Hello, [here](https://github.com/FeiGeChuanShu/ncnn_Android_face/blob/main/ncnn-android-yolov7_face/app/src/main/jni/yoloface.cpp#L281) passing 0 for mean values lead to skipping of mean subtraction step. However before that in code pass mean values equal to 127.5f. So which is correct?

gordinmitya

大佬，感谢分享。 跑下来就是面部分割有些问题，有时候会全黑。 但是再跑可能会有结果出来。 不知道如何解决

zhaoyk1986