DefTruth

I have converted the model to ONNX and implemented a C ++ interface based on onnxruntime. see [colorizer.cpp](https://github.com/DefTruth/litehub/blob/main/ort/cv/colorizer.cpp)

opencv conflict, not user friendly

1

## 🐛 Bug I already have opencv 4.5.1 in my device, but albumentations still try install opencv-python-headless into my device, this make conflict happens between two installed opencv ! how...

opencv conflict, not user friendly

2

## 🐛 Bug I already have opencv 4.5.1 in my device, but albumentations still try install opencv-python-headless into my device, this make conflict happens between two installed opencv ! how...

👉How to add your own model (如何添加你的模型？)

14

这个issue主要讲一下，如何把你自己的模型添加到lite.ai.toolkit。lite.ai.toolkit集成了一些比较新的基础模型，比如人脸检测、人脸识别、抠图、人脸属性分析、图像分类、人脸关键点识别、图像着色、目标检测等等，可以直接用到具体的场景中。但是，毕竟lite.ai.toolkit的模型还是有限的，具体的场景下，可能有你经过优化的模型，比如你自己训了一个目标检测器，可能效果更好。那么，如何把你的模型加入到lite.ai.toolkit中呢？这样既能用到lite.ai.toolkit一些已有的算法能力，也能兼容您的具体场景。这个issue主要是讲这个问题。大家有疑惑的可以提在这个issue，我会尽可能回答~

[FAQ]: 作者回答👉关于如何根据抠图模型的alpha合成新背景？

1

由于问这个问题的人比较多，我在lite里面增加了一些辅助函数来实现背景合成，但还没合并进主分支，具体细节可以参考以下这段逻辑，有需要的可以参考一下（目前不考虑性能的优化问题，有需要的同学可以自己根据这段逻辑做特定的性能优化）： ```c++ void lite::utils::swap_background(const cv::Mat &fgr_mat, const cv::Mat &pha_mat, const cv::Mat &bgr_mat, cv::Mat &out_mat, bool fgr_is_already_mul_pha) { // user-friendly method for background swap. if (fgr_mat.empty() || pha_mat.empty() || bgr_mat.empty()) return;...

An onnxruntime c++ implementation for UltraFace. See [ultraface.cpp](https://github.com/DefTruth/litehub/blob/main/ort/cv/ultraface.cpp) .

👉An ONNXRuntime C++ Demo for RobustVideoMatting is released.

4

* ONNXRuntime C++ Demo: [lite.ai.toolkit](https://github.com/DefTruth/lite.ai.toolkit/blob/main/ort/cv/rvm.cpp) from [DefTruth](https://github.com/DefTruth) * [rvm.cpp](https://github.com/DefTruth/lite.ai.toolkit/blob/main/ort/cv/rvm.cpp) * [rvm.h](https://github.com/DefTruth/lite.ai.toolkit/blob/main/ort/cv/rvm.h) * [rvm-cpp-demo](https://github.com/DefTruth/lite.ai.toolkit/blob/main/examples/lite/cv/test_lite_rvm.cpp) * the output using rvm_mobilenetv3_fp32 of [rvm-cpp-demo](https://github.com/DefTruth/lite.ai.toolkit/blob/main/examples/lite/cv/test_lite_rvm.cpp) is:   * 详细的RobustVideoMatting C++版本接口使用文档见 [RobustVideoMatting.lite.ai.toolkit](https://github.com/DefTruth/RobustVideoMatting.lite.ai.toolkit)

有没有coco上训好的模型呢？

4

有没有coco上训好的模型呢？

add YOLOX, YOLOR and Scaled-YOLOv4

1

* Features: * [YOLOX 2021 🔥🔥🔥new↑](https://github.com/Megvii-BaseDetection/YOLOX) * [YOLOR 2021 🔥🔥new↑](https://github.com/WongKinYiu/yolor) * [Scaled-YOLOv4 CVPR 2021 🔥🔥🔥↑](https://github.com/WongKinYiu/ScaledYOLOv4) @amusi

## add MNN/TNN/ONNXRuntime C++ demo: * [Photo2Cartoon MNN C++](https://github.com/DefTruth/lite.ai.toolkit/blob/main/lite/mnn/cv/mnn_female_photo2cartoon.cpp) * [Photo2Cartoon TNN C++](https://github.com/DefTruth/lite.ai.toolkit/blob/main/lite/tnn/cv/tnn_female_photo2cartoon.cpp) * [Photo2Cartoon ONNXRuntime C++](https://github.com/DefTruth/lite.ai.toolkit/blob/main/lite/ort/cv/female_photo2cartoon.cpp) ## usage ```C++ #include "lite/lite.h" static void test_default() { std::string head_seg_onnx_path = "../../../hub/onnx/cv/minivision_head_seg.onnx";...