Research map / Video / Action recognition
More Is Less: Learning Efficient Video Representations by Big-Little Network and Depthwise Temporal Aggregation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add learned TAM temporal aggregation to TSN (ablation) | TSN with ResNet-50 · SS-V1 | Top-1 accuracy: 17.4 -> 46.1 | Helped |
| Add TAM instead of TSM global fusion (ablation) | TSN with ResNet-50 · SS-V1 | Top-1 accuracy: 43.4 -> 46.1 | Helped |
| Add local fusion by merging branches processing adjacent frames (ablation) | bLVNet with bLResNet-50 · SS-V1 | Top-1 accuracy: 17.4 -> 33.6 | Helped |
| Add TAM global fusion to local fusion (ablation) | bLVNet with bLResNet-50 · SS-V1 | Top-1 accuracy: 33.6 -> 46.4 | Helped |
| Use TAM rather than TSM for global fusion (ablation) | bLVNet with bLResNet-50 · SS-V1 | Top-1 accuracy: 44.2 -> 46.4 | Helped |
| Increase the number of input frames (ablation) | bLVNet-TAM · SS-V1 | Top-1 accuracy | Helped |
| Increase input frames from 8×2 to 16×2 (ablation) | bLVNet-TAM with bLResNet-50 · Moments-in-time | Top-1 accuracy: 31.2 -> 31.4 | Helped |
| Use bLVNet-TAM instead of TSN | Video action recognition · same number of input frames | memory usage: ∼2 times | Helped |
Closest papers by meaning
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · similarity 0.90)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.89)
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.89)
- Look More but Care Less in Video Recognition (NeurIPS 2022 · similarity 0.89)
- Compressed Video Action Recognition (CVPR 2018 · similarity 0.89)
- Recognize Actions by Disentangling Components of Dynamics (CVPR 2018 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.