Research map / Video / Action recognition
Structured Label Inference for Visual Understanding
Nelson Nauata, Hexiang Hu, Guang-Tong Zhou, Zhiwei Deng, Zicheng Liao, Greg Mori
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 structured label relations (SINN) to the single-frame CNN | single-frame CNN · THUMOS | mAP_L: 35.5 -> 38.2 | Improved |
| add structured label relations (SINN) to the single-frame CNN | single-frame CNN · MultiTHUMOS | mAP_L: 24.6 -> 27.0 | Improved |
| add temporal dependencies with an LSTM | single-frame CNN · THUMOS | mAP_L: 35.5 -> 41.8 | Improved |
| add structured labels and temporal label propagation (siLSTM) to LSTM | LSTM · THUMOS | mAP_L: 41.8 -> 43.9 | Improved |
| add structured labels and temporal label propagation (siLSTM) to LSTM | LSTM · MultiTHUMOS | mAP_L: 30.5 -> 31.5 | Improved |
| use structured inference (SINN) instead of bidirectional inference only (BINN) (ablation) | single-frame model · THUMOS | mAP_L: 37.6 -> 38.2 | Improved |
| add SINN to CNN + Logistics | CNN + Logistics · YouTube-8M | mAP_V: 27.98 -> 31.18 | Improved |
| use SINN instead of BINN (ablation) | CNN features · YouTube-8M | mAP_V: 30.17 -> 31.18 | Improved |
| add partially observed vertical labels to SINN (ablation) | SINN · YouTube-8M | mAP_V: 31.18 -> 57.71 | Improved |
| add audio features to SINN (ablation) | SINN RGB · YouTube-8M V2 | mAP_V: 40.19 -> 42.32 | Improved |
| use SINN instead of Logistics with RGB and audio | logistic baseline · YouTube-8M V2 | mAP_V: 38.61 -> 42.32 | Improved |
| use SINN instead of CNN + Logistics | CNN + Logistics · NUS-WIDE | mAP_L: 46.94 -> 67.20 | Improved |
Closest papers by meaning
- Learning Structured Inference Neural Networks With Label Relations (CVPR 2016 · similarity 0.89)
- Every Moment Counts: Dense Detailed Labeling of Actions in Complex Videos (IJCV 2017 · similarity 0.88)
- Learning from Weakly-Labeled Web Videos via Exploring Sub-concepts (AAAI 2022 · similarity 0.88)
- Temporal Dynamic Graph LSTM for Action-Driven Video Object Detection (ICCV 2017 · similarity 0.88)
- An Effective-Efficient Approach for Dense Multi-Label Action Detection (IJCV 2026 · similarity 0.88)
- Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition (CVPR 2016 · similarity 0.88)
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.
