Matěj Šmíd

A Python PyQt desktop application for tracking insects, small fishes, or bugs. The app enables biologists and ecologists to analyze the behavior of laboratory animals.

I took over a large and messy codebase, refactored the majority of the code to be maintainable and well structured, and contributed new features. The application uses deep learning to locate, track, and re-identify multiple individuals in videos.

This project revolved around an invention driven by curiosity. We were investigating the nature of artifacts in video recordings of an ice hockey match from multiple views. We came to the conclusion that the artifacts were caused by strong photographic flashes distorted by the rolling shutter image sensors in cameras. We quickly found out that this can be used to synchronize an arbitrary number of cameras with very high precision.

I built a data processing library for quality inspection using optical coherence tomography volumetric data. The measurement device was built to inspect a type of material for the reaction chamber of a tokamak fusion reactor.

I developed a C++ application for tracking cancer cells in time-lapse videos and a Python suite for analysis of cell motion patterns. The goal was to quantify the influence of anti-cancer drugs that reduce cell motility.

I have adapted state-of-the-art multiple object tracking algorithms to track floorball players using eight cameras.

The method demonstrated in the video uses probabilistic occupancy maps algorithm to fuse data from multiple views and K-shortest paths graph algorithm to find trajectories in the 2D-occupancy maps.