I extremely optimized OpenTLD, all useless contents were removed. This time, tracking speed is about 3ms. Besides, I deleted initialization funtion and OpenCV 3.x is supported now.
I optimized OpenTLD making it run faster and better for face tracking.
This version of TLD below is faster and more stable than that in OpenCV. I delete some funtions to make it run faster. What is more, use RNet to judge the face that TLD produced to avoid TLD tracking a wrong target. In order to get a stable bounding box, I fix the width and height that MTCNN provides when initialising. Running speeds on my PC(Intel® Xeon(R) CPU E5-2673 v3 @ 2.40GHz × 48) are about 16ms(MTCNN, ncnn), 30ms(TLD initialization), 10ms(TLD tracking) for an image in 320*240 resolutions. Besides, MTCNN can be replaced by PCN or any other face detection algorithms.
中文地址:https://blog.csdn.net/renhanchi/article/details/85089265
OpenCV 2.4.X is required!(Now OpenCV 3.x supported)
Reset your ncnn path in CMakeLists.txt.
mkdir build
cd build
cmake ..
make
cd ..
./demo