This is an implementation of the [Viola and Jones Object Detection Framework][1] in C++. It has been particularly optimized for the face detection paradigm:
- Horizontal flipping face sample images in training phase.
- Rotating negative sample images in training phase.
- 4 Haar features implemented.
- Integral image for faster processing.
- Post-normalization speedup.
- Scale invariant (feature scaling).
Installation --------------
git clone https://github.com/alexdemartos/ViolaAndJones.git violaandjones
cd violaandjones/src
g++ -std=c++0x -lpng -O3 Feature.cpp WeakClassifier.cpp StrongClassifier.cpp CascadeClassifier.cpp main.cpp -o vandj `libpng-config --ldflags`
chmod +x vandj- Run
./vandjwithout arguments to see the options. - No classification models are provided for object/face detection, you must train your own.
- No images nor training samples are provided.
I've used this program to train a face detection algorithm. Here are some results:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent