What Makes Training Multi-Modal Classification Networks Hard?

CVPR 2020  ·  Wei-Yao Wang, Du Tran, Matt Feiszli ·

Consider end-to-end training of a multi-modal vs. a single-modal network on a task with multiple input modalities: the multi-modal network receives more information, so it should match or outperform its single-modal counterpart. In our experiments, however, we observe the opposite: the best single-modal network always outperforms the multi-modal network. This observation is consistent across different combinations of modalities and on different tasks and benchmarks. This paper identifies two main causes for this performance drop: first, multi-modal networks are often prone to overfitting due to increased capacity. Second, different modalities overfit and generalize at different rates, so training them jointly with a single optimization strategy is sub-optimal. We address these two problems with a technique we call Gradient Blending, which computes an optimal blend of modalities based on their overfitting behavior. We demonstrate that Gradient Blending outperforms widely-used baselines for avoiding overfitting and achieves state-of-the-art accuracy on various tasks including human action recognition, ego-centric action recognition, and acoustic event detection.

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Action Classification Kinetics-400 G-Blend Acc@1 77.7 # 122
Action Classification Kinetics-400 G-Blend (Sports-1M pretrain) Acc@1 78.9 # 109
Action Recognition In Videos miniSports G-Blend Clip Hit@1 49.7 # 1
Video hit@1 62.8 # 1
Video hit@5 85.5 # 1
Action Recognition In Videos Sports-1M G-Blend Video hit@1 74.8 # 1
Video hit@5 92.4 # 1


No methods listed for this paper. Add relevant methods here