This is official implementation of paper VidConv: A modernized 2D ConvNet for Efficient Video Recognition
- The paper is under review, and the full Code will be released upon the acceptance decision.
- The tiny version of VidConv backbone, which we used to secure the 3rd place in AI City Challenge 2022 -Track 3: Driver Action Recognition, can be found in https://github.com/cybercore-co-ltd/AICity2022-Track3