WildTrack, in collaboration with the UC Berkeley I School, has won the coveted Hal R. Varian Capstone award for the WildTrackAI Computer Vision application! WildTrack was voted by the judges as the best project out of 22 competitors showcasing AI applications in a wide range of fields.
The WildTrackAI system provides a rapid classification for footprints from endangered species allowing users to upload images to the cloud and get a rapid classification on species and individual. Now in Phase 2, we plan to take this forward to help us process larger volumes of data and complement our FIT in JMP software, for high-level classification, more species and more biometrics!
Some comments from the judges of the competition included:
- The completeness and overall polish of WildTrackAI, that you have implemented in conjunction with WildTrack and released it ‘into the wild,’ and that it is thorough as a ‘zero to one’ project even in this early stage put it ahead for the judges. Impressive work top to bottom.
- The judges felt that this was a unique, inspirational project: in terms of project and problem space, in terms of potential social good, and in terms of execution from technical and project management/implementation perspectives. We feel that this is an excellent representation of the I School’s ideals, and a great example of how data science can contribute to and expand upon the efforts of people and organizations who are working to protect and conserve our world.
- Well-defined problem statement, solution, and long term scalability plan. The selection of toolset, and storytelling in the presentation were also strengths.
- Technical proficiency: we were impressed with the incorporation of a feedback loop on re-training the model, and the end-to-end data analysis pipeline (from user-contributed images to front-facing summary statistics and usable features by administrators).
- Thoughtful about different users and use cases, and some excellent next steps identified as part of the project.
- Excellent that you have partnered with WildTrack and brought in their expertise and guidance to properly define the problem, and to gain an understanding of current data availability, common practices to design around, and what would constitute success, what would help move the needle for WildTrack. It’s clear that you have established a great partnership.
- Evaluation and accuracy checks were quite compelling.