Research map / Video / Action recognition
CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos
Check 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 |
|---|---|---|---|
| Use CDC frame-level predictions for per-frame action labeling | C3D-based localization network · THUMOS'14 per-frame labeling | mAP: 41.7 -> 44.4 | Helped |
| Use CDC frame-level predictions to refine proposal boundaries | THUMOS'14 proposal-based localization · THUMOS'14 temporal localization, IoU 0.5 | mAP: 19.0 -> 23.3 | Helped |
| Replace separate convolution and de-convolution with the joint CDC filter (ablation) | C3D-based per-frame labeling · THUMOS'14 per-frame labeling | mAP: 41.7 -> 44.4 | Helped |
| Train the 3D ConvNet layers end-to-end instead of keeping them fixed (ablation) | CDC · THUMOS'14 per-frame labeling | mAP: 37.4 -> 44.4 | Helped |
| Train the 3D ConvNet layers end-to-end instead of keeping them fixed (ablation) | CDC · THUMOS'14 temporal localization, IoU 0.5 | mAP: 20.4 -> 23.3 | Helped |
| Refine segments using CDC frame-level scores | Wang and Tao predictions · ActivityNet 2016 validation | mAP at IoU 0.75: 4.1 -> 26.0 | Helped |
| Refine segments using CDC frame-level scores | Wang and Tao predictions · ActivityNet 2016 validation | Average-mAP: 16.4 -> 23.8 | Helped |
| Train on the validation set in addition to the training set (ablation) | CDC · ActivityNet 2016 test | Average-mAP: 22.9 -> 22.9 | No clear effect |
| Vary the proposal extension percentage from alpha 1/8 to 1/4 (ablation) | CDC boundary refinement · THUMOS'14 temporal localization, IoU 0.5 | mAP: 23.3; 23.2; 23.1; 23.1; 23.6 | No clear effect |
| Increase temporal prediction granularity from one label per eight frames to one label per frame (ablation) | CDC · THUMOS'14 temporal localization, IoU 0.5 | mAP | Helped |
Closest papers by meaning
- Temporal Action Localization in Untrimmed Videos via Multi-Stage CNNs (CVPR 2016 · similarity 0.93)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.92)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.92)
- Exploring Temporal Preservation Networks for Precise Temporal Action Localization (AAAI 2018 · similarity 0.91)
- Rethinking the Faster R-CNN Architecture for Temporal Action Localization (CVPR 2018 · similarity 0.91)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.91)
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.