One of the most successful and popular feature extraction algorithms used in Computer Vision for object detection is the “Histograms of oriented gradients for human detection” (HOG) by Dalal and Triggs. And one of the best implementations (in terms of speed, reliability, etc.) is by Piotr Dollar and it is available in the “Piotr’s Computer Vision Matlab Toolbox“. The core of the code is written in very compact and efficient C programming language and the interfacing is given in the MATLAB programming language. However, when implementing object detection, activity recognition or other computer vision systems in Java which is currently the most popular programming language and widely used in industry and business, there is no reliable HOG feature extraction API. This project tackles this problem by contributing a Java API to the core C code. Java Native Interface (JNI) is used to bridge C and Java.
Part of the code is available at: https://github.com/Kyaw-Kyaw-Htike/Java-API-for-Histograms-of-Oriented-Gradients-C-code
- Programming languages: Java, C, MATLAB
- Special Expertise Required: Java Native Interface (JNI), Computer Vision
- Contribution: Entire project