Research map / Video / Action recognition
R-C3D: Region Convolutional 3D Network for Temporal Activity Detection
Huijuan Xu, Abir Das, Kate Saenko
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 the reverse pass (two-way buffer) for training data (ablation) | R-C3D · THUMOS'14 | mAP@0.5 (%): 27.0 -> 28.9 | Improved |
| add the reverse pass (two-way buffer) for training data (ablation) | R-C3D · THUMOS'14 mAP@0.1 | mAP@0.5 (%): 51.6 -> 54.5 | Improved |
| use end-to-end R-C3D instead of best prior method | R-C3D · THUMOS'14 | mAP@0.5 (%): 23.3 -> 28.9 | Improved |
| use R-C3D instead of UPC (C3D features only) | R-C3D · ActivityNet validation | mAP@0.5 (%): 22.5 -> 26.8 | Improved |
| use R-C3D instead of UPC (C3D features only) | R-C3D · ActivityNet test | mAP@0.5 (%): 22.3 -> 26.8 | Improved |
| train on train+val instead of train only (ablation) | R-C3D · ActivityNet test | mAP@0.5 (%): 26.8 -> 28.4 | Improved |
| train on train+val instead of train only (ablation) | R-C3D · ActivityNet test average mAP | average mAP (%): 13.1 -> 16.7 | Improved |
| use R-C3D instead of methods with handcrafted features | R-C3D · ActivityNet test | mAP@0.5 (%): 28.4 vs 36.4 | Hurt |
| use R-C3D instead of asynchronous temporal fields (standard evaluation) | R-C3D · Charades | mAP (%): 9.6 -> 12.4 | Improved |
| use R-C3D instead of asynchronous temporal fields (post-processed evaluation) | R-C3D · Charades | mAP (%): 12.1 -> 12.7 | Mixed |
| vary anchor segment scales (ablation) | R-C3D · Charades | mAP (%) | No clear effect |
| share conv features between proposal and classification subnets, versus LSTM-based DAP | R-C3D · inference speed | FPS: 134.1 -> 569 FPS | Improved |
Later papers that compared with this method
9 later accepted papers report R-C3D as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.
- Three Birds with One Stone: Multi-Task Temporal Action Detection via Recycling Temporal Annotations (CVPR 2021 · better on 2 of 2 shared results)
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · better on 2 of 2 shared results)
- Graph Attention Based Proposal 3D ConvNets for Action Detection (AAAI 2020 · better on 5 of 5 shared results)
- Representation Learning on Visual-Symbolic Graphs for Video Understanding (ECCV 2020 · better on 1 of 1 shared result)
- Comprehensive Instructional Video Analysis: The COIN Dataset and Performance Evaluation (TPAMI 2020 · better on 1 of 1 shared result)
- COIN: A Large-Scale Dataset for Comprehensive Instructional Video Analysis (CVPR 2019 · better on 1 of 1 shared result)
- Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognition (NeurIPS 2019 · better on 1 of 1 shared result)
- Video Imprint Segmentation for Temporal Action Detection in Untrimmed Videos (AAAI 2019 · better on 1 of 1 shared result)
1 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Two-Stream Region Convolutional 3D Network for Temporal Activity Detection (TPAMI 2019 · similarity 0.97)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.90)
- Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification (ECCV 2018 · similarity 0.90)
- CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (CVPR 2017 · similarity 0.90)
- Efficient Activity Detection in Untrimmed Video with Max-Subgraph Search (TPAMI 2016 · similarity 0.89)
- ZSTAD: Zero-Shot Temporal Activity Detection (CVPR 2020 · similarity 0.89)
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.
