Research map / Video / Action recognition
MiniROAD: Minimal RNN Framework for Online Action Detection
Joungbin An, Hyolim Kang, Su Ho Han, Ming-Hsuan Yang, Seon Joo Kim
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 |
|---|---|---|---|
| use step-function loss weights (loss only at last clip step) instead of uniform weights (ablation) | MiniROAD GRU · THUMOS'14 | mAP (%): 66.4 -> 71.8 | Improved |
| use step-function loss weights instead of learnable weights (ablation) | MiniROAD GRU · THUMOS'14 | mAP (%): 70.8 -> 71.8 | Improved |
| linearly increasing instead of uniform loss weights (ablation) | MiniROAD GRU · THUMOS'14 | mAP (%): 66.4 -> 68.6 | Improved |
| apply proposed last-step loss weights to the transformer encoder (ablation) | OadTR encoder · THUMOS'14 | mAP (%): 63.0 -> 64.9 | Improved |
| increase training clip length beyond 32 seconds (ablation) | MiniROAD · THUMOS'14 | mAP (%) | No clear effect |
| GRU with step weights versus transformer SOTA on FineAction | MiniROAD · FineAction | mAP (%): 31.8 -> 37.1 | Improved |
| use faster NVIDIA optical flow instead of TV-L1 | MiniROAD · THUMOS'14 | Overall FPS / mAP: 23.8 -> 122 FPS | Mixed |
| apply last-step loss to action anticipation | MiniROAD · THUMOS'14 anticipation (Kinetics) | average mAP: 56.8 -> 59.7 | Improved |
Later papers that compared with this method
3 later accepted papers report MiniROAD as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Backtrace Mamba: Reviving Critical Temporal Contexts via Hierarchical Memory Compression for Online Action Detection (AAAI 2026 · better on 1 of 1 shared result)
- Hierarchical Event Memory for Accurate and Low-latency Online Video Temporal Grounding (ICCV 2025 · better on 3 of 3 shared results)
- Open-ended Hierarchical Streaming Video Understanding with Vision Language Models (ICCV 2025 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GateHUB — GateHUB: Gated History Unit With Background Suppression for Online Action Detection (CVPR 2022) · better on 2 of 3 shared results
- OadTR — OadTR: Online Action Detection With Transformers (ICCV 2021) · better on 1 of 1 shared result
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
- Context-Enhanced Memory-Refined Transformer for Online Action Detection (CVPR 2025 · similarity 0.89)
- OadTR: Online Action Detection With Transformers (ICCV 2021 · similarity 0.88)
- Online Action Detection (ECCV 2016 · similarity 0.88)
- WOAD: Weakly Supervised Online Action Detection in Untrimmed Videos (CVPR 2021 · similarity 0.88)
- E2E-LOAD++: End-to-end Long-form Online Action Detection and Beyond (TPAMI 2026 · similarity 0.87)
- Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection? (NeurIPS 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.
