Here you will find some of the stuff I am working on at Aimetis. Aimetis makes various video analytics products for different applications such as tracking people, cars, counting people and detecting left/removed items. Most of the algorithms below were developed by me, however some algorithms were modified or improved by myself and they will be indicated.

Main challenge at Aimetis is to make robust algorithms that can handle various working environment, yet use very little CPU. To meet our CPU usage requirements, on average, an algorithms needs to process a CIF frame (320x240 pixels) in about 15 ms - this is approximately 66 frames per second (FPS). This might seem overkill because we never need to analyze a video at 66 FPS; we typically do analysis at 5 FPS. However, keep in mind one CPU needs to run multiple cameras as well decompress and recompress video feeds from the cameras.

Here are some of the results we are getting on our algorithms. For the obvious reasons, I am not at liberty to discuss the technical details of the algorithms. You can download the entire video set as mpeg files here or you can see flash versions categorized by the algorithm type.