Research map / Video / Action recognition
Learning Temporal Action Proposals With Fewer Labels
Jingwei Ji, Kaidi Cao, Juan Carlos Niebles
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 Time Warping perturbation (ablation) | Mean Teacher with Time Masking · THUMOS14 (60% labels) | AR@100: 44.79 -> 46.87 | Improved |
| add Time Masking perturbation (ablation) | Mean Teacher with Time Warping · THUMOS14 (60% labels) | AR@100: 45.37 -> 46.87 | Improved |
| add Time Masking instead of regular dropout (ablation) | Mean Teacher · THUMOS14 (60% labels) | AR@100: 44.25 -> 46.87 | Improved |
| add regular dropout to Mean Teacher (ablation) | Mean Teacher · THUMOS14 (60% labels) | AR@100: 44.20 -> 44.25 | No clear effect |
| add VAT to Mean Teacher (ablation) | Mean Teacher · THUMOS14 (60% labels) | AR@100: 44.20 -> 44.21 | No clear effect |
| add Mean Teacher semi-supervised training with unlabeled videos | BSN · THUMOS14 (60% labels) | AR@100: 40.79 -> 44.20 | Improved |
| add VAT adversarial noise | BSN · THUMOS14 (60% labels) | AR@100: 40.79 -> 43.13 | Improved |
| use semi-supervised framework with 60% labels instead of fully supervised BSN with 60% labels | BSN · THUMOS14 | AR@100: 40.79 -> 46.87 | Improved |
| semi-supervised with 60% labels vs BSN with all labels | BSN · ActivityNet v1.3 | AR@100: 74.16 -> 75.07 | Improved |
| semi-supervised with 60% labels vs BSN with all labels | BSN · THUMOS14 | AR@100: 46.06 -> 46.87 | Improved |
| add unlabeled-data training when 100% labels available | BSN · THUMOS14 | AR@100: 46.06 -> 47.53 | Improved |
| use semi-supervised proposals for localization | BSN + UNet · THUMOS14 test | mAP@0.7: 20.0 -> 20.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CTAP — CTAP: Complementary Temporal Action Proposal Generation (ECCV 2018) · better on 2 of 2 shared results
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
- Temporal Action Detection With Multi-Level Supervision (ICCV 2021 · similarity 0.92)
- Action Sets: Weakly Supervised Action Segmentation Without Ordering Constraints (CVPR 2018 · similarity 0.91)
- Semi-Supervised Action Recognition With Temporal Contrastive Learning (CVPR 2021 · similarity 0.91)
- Semi-Supervised Temporal Action Detection with Proposal-Free Masking (ECCV 2022 · similarity 0.91)
- Self-Supervised Learning for Semi-Supervised Temporal Action Proposal (CVPR 2021 · similarity 0.90)
- Stable Mean Teacher for Semi-supervised Video Action Detection (AAAI 2025 · 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.
