Skip to content

mediapipe-hand,mediapipe-body,mediapipe-face, mediapipe-embedding, mediapipe-classifier and so on.MNN inference

Notifications You must be signed in to change notification settings

MirrorYuChen/Mediapipe-MNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

59 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mediapipe mnn inference

  • Author: chenjingyu 20230619

1.项目说明

  • mediapipe模型推理测试代码;

2.更新日志

时间 更新内容
20230619 1.创建项目;2.添加相关依赖库;
20230620 1.添加相关转换后mnn模型;
20230625 1.添加手掌检测相关推理代码;
20230626 1.添加手掌关键点模型相关推理代码;
20230727 1.添加推理测试代码;
20230729 1.添加人脸检测相关模型及基本推理代码;2.添加人脸检测相关解析代码及测试代码;3.添加Blazeface的大模型推理代码;
20230730 1.添加Mediapipe的face mesh推理代码;2.添加MNN的人脸检测模型推理代码;
20230802 1.添加最新版本的YuNet推理代码;
20230814 1.添加image embedding相关推理代码;
20230819 1.添加Knift模型推理代码;
20230820 1.添加BlazePose人体检测模型及推理代码;2.添加BlazePose人体关键点检测模型及推理代码;3.添加图像分类模型及推理代码;
20230822 1.添加mediapipe的selfie_segmentation模型推理demo;

3.一些效果

示例 效果图
mediapipe-classifier mediapipe-classifier
mediapipe-face mediapipe-face
mediapipe-hand mediapipe-hand
mediapipe-pose mediapipe-pose
mediapipe-knift mediapipe-knift
mediapipe-embbeding mediapipe-embedding
mediapipe-selfie_segmentation mediapipe-selfie_segmentation
YN-facedetection YN-facedetection
Lite-facedetection Lite-facedetection
  • 实测knift效果很差,大概率是我这边接的有问题,但是没发现是什么问题,脑壳疼!后面再研究一下mediapipe源码看看。

4.参考项目

5.可视化网络输入

#include <opencv2/opencv.hpp>

using namespace MNN;
CV::ImageProcess::Config image_process_config;
image_process_config.filterType = CV::BILINEAR;
image_process_config.sourceFormat = CV::BGR;
image_process_config.destFormat = CV::BGR;
image_process_config.wrap = CV::ZERO;
std::shared_ptr<CV::ImageProcess> pretreat =
  std::shared_ptr<CV::ImageProcess>(CV::ImageProcess::create(image_process_config));
pretreat->setMatrix(trans_);

cv::Mat dst_image = cv::Mat(input_h_, input_w_, CV_8UC3);
std::shared_ptr<MNN::Tensor> dst_tensor(MNN::Tensor::create<uint8_t>(std::vector<int>{1, dst_image.rows, dst_image.cols, dst_image.channels()}, dst_image.data));
pretreat->convert((uint8_t *)in.data, in.width, in.height, in.width_step, dst_tensor.get());
cv::imshow("result", dst_image);
cv::waitKey(0);

About

mediapipe-hand,mediapipe-body,mediapipe-face, mediapipe-embedding, mediapipe-classifier and so on.MNN inference

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published