My goal is to transfer my skills, knowledge, and various tricks of the trade I've learned over the years while working in the field of applied computer vision. I am planning to build and iterate on a set of python packages (potentially with some minimal C/C++ ports) solving some common problems in the field. My experience covers building custom object detectors, single and multi-object trackers, metric learning systems, and various productivity tools that can improve the process of models iteration, datasets curation, interactive CV applications debugging and tuning (and more). I believe that strong baselines of such components can help other people build the core functionalities of their projects without being distracted by the typical side-quests.

My first goal is to restore and improve the not-so-up-to-date motpy package. so that it can be treated as a great baseline for incorporating multi-object tracking algorithm in any application that requires such a capability. I want it to be as parameter-less as possible, while giving the users lots of room for tuning and customization of the pipeline, utilized motion models and generally the tracker management behavior. At the same time, I want to make sure it has significantly better debugging capabilities, while still maintaining the easiness of use out of the box.

The next projects will be most likely focused on one of the following: an object detection auto-annotation toolkit or a python package with various visualization and debugging tools that I find very useful in all kinds of CV projects.

Personally, I am a CV/DL/ML engineer with more than 10 years of experience in the field. I am currently living in Warsaw, Poland :) I am planning to use the funds to buy myself a coffee, and as a proxy motivator for spending more time on open source projects.