Video summarization to facilitate browsing of a large database of underwater videos

Project Statement

Using innovative computer vision techniques, create a single still image which is a mosaic of all the organisms visible in a full-length underwater video.


Fixed location video cameras deployed on ocean observatories convey live imagery from the ocean depths to the shore, which is used extensively for biological research. Ocean Networks Canada hosts video cameras in diverse regions, from the coast to the abyss to the Arctic, which record hundreds of hours of video every week. Researchers browsing the data archive need a way to see previews of the highlights of a given video segment before committing to downloading it. Biologists are typically interested in the organisms, whether moving or stationary, in the video. Analysis of underwater video poses unique challenges to existing techniques in computer vision, including, but not limited to turbidity, limited and uneven lighting, moving particles in the water and inaccessibility of environment.

Relevant data

Skills Required

Programming skills in Matlab, Python or other language as applicable, advanced computer vision, knowledge of deep-sea organisms an asset.