Research map / Video / Video segmentation and tracking
Attentive Temporal Pyramid Network for Dynamic Scene Classification
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 |
|---|---|---|---|
| replace average pooling with pyramid attention matrix (N=7, M=32) (ablation) | ResNet-50 features · YUPENN++ | accuracy (%): 89.53 -> 92.03 | Helped |
| replace average pooling with pyramid attention matrix (N=7, M=32) (ablation) | ResNet-50 features · Maryland | accuracy (%): 89.23 -> 93.85 | Helped |
| increase pyramid subsets N from 1 to 7 (M=32) (ablation) | attention matrix · Maryland | accuracy (%): 90.77 -> 93.85 | Helped |
| increase attention modules M from 1 to 32 (N=7) (ablation) | attention matrix · YUPENN++ | accuracy (%): 89.63 -> 92.03 | Helped |
| increase N and M further (ablation) | attention matrix · YUPENN++ | accuracy (%): 92.03 -> 92.03 (N=3 vs 7, M=32) | No clear effect |
| use a single attention module (N=1, M=1) (ablation) | attention matrix · YUPENN++ | accuracy (%): 89.53 -> 88.05 | Hurt |
| replace concatenation with kernel aggregation layer (ablation) | ATP-Net · Maryland | accuracy (%): 90.77 -> 95.38 | Helped |
| replace FC layer with kernel aggregation layer (ablation) | ATP-Net · Maryland | accuracy (%): 93.85 -> 95.38 | Helped |
| replace FC layer with kernel aggregation layer (ablation) | ATP-Net · YUPENN++ | accuracy (%): 92.03 -> 92.37 | Helped |
| ATP-Net vs ResNet-50 average pooling | ResNet-50 · ActivityNet val | accuracy (%): 65.3 -> 74.6 | Helped |
| ATP-Net (ResNet-50) vs P3D ResNet (ResNet-152) | ATP-Net · ActivityNet val | accuracy (%): 75.1 -> 74.6 | Hurt |
Closest papers by meaning
- Temporal Residual Networks for Dynamic Scene Recognition (CVPR 2017 · similarity 0.89)
- Temporal Pyramid Network for Action Recognition (CVPR 2020 · similarity 0.86)
- Fine-Grained Dynamic Head for Object Detection (NeurIPS 2020 · similarity 0.85)
- Spatiotemporal Pyramid Network for Video Action Recognition (CVPR 2017 · similarity 0.85)
- DyST: Towards Dynamic Neural Scene Representations on Real-World Videos (ICLR 2024 · similarity 0.84)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · similarity 0.83)
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.