VIPriors 2: Visual Inductive Priors for Data-Efficient Deep Learning Challenges
This work addresses data scarcity issues in computer vision for researchers and practitioners, but it is incremental as it builds on prior challenges and focuses on established techniques.
The paper tackled the problem of data-efficient deep learning by organizing challenges where models were trained from scratch on limited data for computer vision tasks, resulting in baselines being outperformed by a large margin across all five challenges.
The second edition of the "VIPriors: Visual Inductive Priors for Data-Efficient Deep Learning" challenges featured five data-impaired challenges, where models are trained from scratch on a reduced number of training samples for various key computer vision tasks. To encourage new and creative ideas on incorporating relevant inductive biases to improve the data efficiency of deep learning models, we prohibited the use of pre-trained checkpoints and other transfer learning techniques. The provided baselines are outperformed by a large margin in all five challenges, mainly thanks to extensive data augmentation policies, model ensembling, and data efficient network architectures.