Junjue Wang


9127 Gates Hillman Center, 4902 Forbes Ave


junjuew dot cs dot cmu dot edu
Github / Twitter

Computer Vision Tools for programmers

For application developers, especially mobile app developers, finding the correct vision library is not an easy task. A usual to-go is OpenCV library, a comprehensive and robust computer vision library that can run on most platforms, including Linux, Windows, Mac OS, Android, and iOS. OpenCV is a SwissTool for computer vision. It includes an army of widely-used computer vision algorithms for simple tasks such as color conversion to complicated processing such as face recognition. However, as a SwissTool, OpenCV can be slow for a lot of computer vision tasks. Particularly, for mobile devices, OpenCV has too much footprint to satisfy the real-time latency requirements. Besides, it may also contain the particular algorithm you want.

Here I put together a list of OpenCV alternatives that can facilitate building you computer vision applications.

Library CPU GPU Support Mobile Oriented Comments
OpenCV yes no cuda api, limited functions  
dlib yes no no  
openVX yes no yes  
openGL no yes    
GPUImage no yes    
Accelerated CV