Research map / Video / Action recognition
Exploring Temporal Preservation Networks for Precise Temporal Action Localization
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 standard 3D convolutions with temporal-preservation filters throughout the network | C3D · THUMOS'14 frame-level | mAP: 44.4 -> 49.5 | Helped |
| Use frame grouping instead of proposal-based refinement | TPC frame-level predictions · THUMOS'14 segment-level, IoU 0.3–0.7 | mAP: TPC + FGM: 44.1, 37.1, 28.2, 20.6, 12.7; TPC + S-CNN: 41.9, 32.5, 25.3, 14.7, 9.0 | Helped |
| Use TPC filters only in conv2 instead of throughout (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 45.5 | Hurt |
| Use TPC filters only in conv3 instead of throughout (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 45.1 | Hurt |
| Use TPC filters only in conv4 instead of throughout (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 45.0 | Hurt |
| Use TPC filters in conv2 and conv3 only (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 46.4 | Hurt |
| Use TPC filters in conv3 and conv4 only (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 45.7 | Hurt |
| Add global average pooling and conv6-GAP (ablation) | TPC · THUMOS'14 frame-level | mAP: 49.5 -> 47.2 | Hurt |
| Add global average pooling and conv6-GAP (ablation) | TPC · THUMOS'14 segment-level, IoU 0.5 | mAP: 25.3 -> 23.6 | Hurt |
| Preserve temporal resolution at all convolution stages rather than only selected stages (ablation) | TPC · THUMOS'14 frame-level and segment-level localization | mAP: Frame-level: 49.5 vs. 45.5, 45.1, 45.0, 46.4, 45.7 | Helped |
Closest papers by meaning
- Temporal Action Localization in Untrimmed Videos via Multi-Stage CNNs (CVPR 2016 · similarity 0.92)
- CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (CVPR 2017 · similarity 0.91)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.90)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.88)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.88)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.86)
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.