Enabling World Class Research Through The Community: Food Recognition Benchmark

The Food Recognition Benchmark has been one of our most cherished projects at AIcrowd. It also stands as a testament to our vision of enabling world class research through the AIcrowd community.

Here we are sharing some exciting updates on how the community collaboration has worked so far and what it’s enabling!

📝 The Paper

The manuscript which describes the first iteration of the Food Recognition Benchmark, and the solutions of the top participants has been accepted for publication. You can check out the pre-print of the manuscript here.

We have tried our best to replicate the top-performing solutions, and have also released the code used for the reproducible experiments reported in the paper.

Also, as in the last rounds, we are looking forward to co-author a paper with the participants providing the future winning solutions. So, if you haven’t, do participate in the Food Recognition Benchmark 2022!

Check out the paper manuscript here | Check out the code repo containing all experiments here

🏷️ Assisted Annotations

The best models from the previous iterations of the benchmark are already in production in the MyFoodRepo app, which is used in multiple research cohorts. This growing dataset will be the basis for future versions of the Food Recognition Benchmark.

In addition, the MyFoodRepo app has recently launched in the first research project in the US, and will launch in research cohorts in three other countries this year.

The Food Recognition Benchmark is not a competition, but an ongoing collaboration with a community of inspiring individuals like yourself, to deliver the future of diet tracking in research cohorts around the world.

Thank you!