This is where all the old stuff from baboon team lives. Due to multiple failed approaches and refactors, a lot of deprecated code was moved to this repo to be archived.
Webscraper to download streamed videos from the San Diego Zoo's Baboon Live Cams. These videos habe been determined to not be useful at the current time in the project, as drone footage has very different properties.
Attempt to track baboons using Kernelized Correlation Filters. (TODO: Insert blurb here to explain how well this worked)
Attempt to detect baboons using simple background subtraction methods. Works well with stationary cameras, but translational and rotational background movement causes issues with this approach.
Interactive test of various common builtin opencv tracking methods. Allows user to draw bounding box around baboon, and attempts to track it for duration of the video. Only able to track, not detect baboons.
Attempt to isolate moving foreground from a variable background by implementing the algorithm described in this paper. At the moment, algorithm works well but is incredibly slow, as it has not been optimized for performance.
Use registration (stabilization) functions from variable_background_tracking, and apply regular background subtraction on the results