One of the first projects I have worked on is my bachelor thesis. In here, I used one of the traffic cameras from the university as input and developed a system to track and count cars using SIFT. Even though I succeed completing the task in the majority of the cases, the bad quality of the camera decreases the quality of the results introducing many of false positives. Read More...

Skin Detector based on the paper: Jones, M.J., Rehg, J.M.: Statistical color models with application to skin detection. IJCV 46(1), 81-96 (2002). We used an RGB-D camera to train and test the models. Code Available! Read More...

We created a set of tools to obtain and save in different formats the data obtained from an RGB-D device, such as Kinect, Primesense cameras, Asus Xtion and others. Code Available! Read More...

This project titled: Calibration and Simultaneous Localization and Mapping, is an extension of the algorithm developed by Jörg Stückler et. al. (MRSMAP). In this case we modified it to work correctly with multiple cameras and to obtain the extrinsic parameters of each of them at the same time it performs SLAM. Also, to improve the results even further, a detector for undefined registrations it is being implemented to avoid errors in scenes with almost no features in depth and color (i.e. white walls) in one of the cameras, that can deteriorate badly the results of the whole system. Read More...

Intrudy is a project done in the lab of mobile robotics in Bonn University. It was developed by Ilya Manyugin and me using ROS and python. The idea of this project is to use a modified Roomba robot with a SICK laser, to detect, track and follow a moving target. Using the laser measurements we are able to detect the movement of a target in a room, to track it, we used ICP on the laser scans to match the one from the previous time step. Then, using this map, we find the places where it changed, and using then we determine where the movement occurred using the amount of laser scans involved and the first derivative of the scans. Code Available! Read More...

A data viewer for the ground truth generated in the scientific paper: "Capturing Hands in Action using Discriminative Salient Points and Physics Simulation" by Tzionas, D. et. al. Read More...