Abstract
The intent of this project is to explore the tracking capabilities of V1KU a neural network system. V1KU is a product by General Vision Company that comprises of CogniMem neural network chip for real-time image learning and CogniSight image recognition engine. The board also consists of Micron/Aptina monochrome CMOS sensor for visual input. The board has powerful capability to learn and recognize objects simultaneously within a fraction of a second. Due to this ability an application is developed which uses board’s capabilities to track a learned object in real-time. The development of this application has gone through various phases of experiments as during initial development stages the board was quite new and very little support was available. After applying the methodology of trial and error I was able to achieve a real-time tracking working with this board. The people at General Vision also gave their inputs on how to optimize the code so that tracking works efficiently. The board has the capabilities to track multiple objects simultaneously, but at this present time the goal is to effectively track a single object. The new version of the board with casing came out recently which has some mounting space that can be utilized in future to mount servo motors to automate the tracking process. The output of this application forms a basis for stereoscopic tracking of various objects in real-time.