Research map / Detection and segmentation / CNN features
Going Deeper into First-Person Activity Recognition
Minghuang Ma, Haoqi Fan, Kris M. Kitani
Check 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 |
|---|---|---|---|
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · GTEA(61)* | activity recognition accuracy: 53.01 -> 75.08 | Improved |
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · GTEA(61) | activity recognition accuracy: 50.45 -> 73.02 | Improved |
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · GTEA(71) | activity recognition accuracy: 47.07 -> 73.24 | Improved |
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · Gaze(25)* | activity recognition accuracy: 28.42 -> 62.40 | Improved |
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · Gaze(40)* | activity recognition accuracy: 16.00 -> 43.42 | Improved |
| jointly fine-tune motion and object streams with action, object, and activity losses instead of SVM fusion (ablation) | two-stream activity recognition · Gaze+(44) | activity recognition accuracy: 34.75 -> 66.40 | Improved |
| add joint motion-object training (ablation) | object recognition · GTEA(71) | object recognition accuracy: 67.74 -> 76.15 | Improved |
| add joint motion-object training (ablation) | object recognition · Gaze(40) | object recognition accuracy: 38.05 -> 55.55 | Improved |
| add joint motion-object training (ablation) | object recognition · Gaze+(44) | object recognition accuracy: 61.87 -> 74.34 | Improved |
| add joint motion-object training (ablation) | action recognition · GTEA(71) | action recognition accuracy: 75.85 -> 78.33 | Improved |
| add joint motion-object training (ablation) | action recognition · Gaze(40) | action recognition accuracy: 33.65 -> 36.27 | Improved |
| add joint motion-object training (ablation) | action recognition · Gaze+(44) | action recognition accuracy: 62.62 -> 65.05 | Improved |
Closest papers by meaning
- First Person Action Recognition Using Deep Learned Descriptors (CVPR 2016 · similarity 0.93)
- Egocentric Action Recognition by Automatic Relation Modeling (TPAMI 2022 · similarity 0.89)
- Cascaded Interactional Targeting Network for Egocentric Video Analysis (CVPR 2016 · similarity 0.89)
- Multi-Dataset, Multitask Learning of Egocentric Vision Tasks (TPAMI 2021 · similarity 0.88)
- Interactive Prototype Learning for Egocentric Action Recognition (ICCV 2021 · similarity 0.88)
- Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learning (ICLR 2025 · 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.
