Research map / Video / Action recognition
Actor and Action Video Segmentation From a Sentence
Kirill Gavrilyuk, Amir Ghodrati, Zhenyang Li, Cees G. M. Snoek
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 Flow to the RGB stream (ablation) | Sentence-guided video segmentation model · A2D Sentences test | Overall IoU: 53.6% -> 55.1% | Improved |
| Use a larger temporal context, increasing input frames from 1 to 16 (ablation) | RGB video encoder · A2D Sentences test | Overall IoU: 48.2% -> 53.6% | Improved |
| Replace the LSTM sentence encoder with a 1D convolutional encoder (ablation) | Sentence-guided video segmentation model · A2D Sentences test | Overall IoU: 51.8% -> 53.6% | Improved |
| Replace the bidirectional LSTM sentence encoder with a 1D convolutional encoder (ablation) | Sentence-guided video segmentation model · A2D Sentences test | Overall IoU: 52.1% -> 53.6% | Improved |
| Add supervision at multiple resolutions rather than only the highest resolution (ablation) | Sentence-guided video segmentation model · A2D Sentences test | Overall IoU: 49.4% -> 53.6% | Improved |
| Use Flow input instead of RGB input (ablation) | Video encoder · A2D Sentences test | Overall IoU: 53.6% -> 49.5% | Hurt |
| Use RGB + Flow fusion instead of RGB only (ablation) | Sentence-guided video segmentation model · A2D Sentences test | mAP: 19.8 -> 21.5 | Improved |
| Use the proposed sentence-guided model rather than the strongest fine-tuned image baseline | A2D Sentences segmentation · A2D Sentences test | Mean IoU: 35.4 -> 42.6 | Improved |
| Use the proposed model rather than Hu et al. | J-HMDB Sentences segmentation · J-HMDB Sentences test | mAP: 17.8 -> 23.3 | Improved |
| Use the proposed model rather than Hu et al. | J-HMDB Sentences segmentation · J-HMDB Sentences test | Overall IoU: 54.6 -> 54.1 | Hurt |
| Use the proposed model rather than Kalogeiton et al. | A2D actor segmentation · A2D | Class-Average accuracy: 73.7 -> 71.4 | Hurt |
| Use the proposed model rather than Kalogeiton et al. | A2D actor and action segmentation · A2D | Mean IoU: 29.7 -> 34.8 | Improved |
Closest papers by meaning
- Context Modulated Dynamic Networks for Actor and Action Video Segmentation with Language Queries (AAAI 2020 · similarity 0.91)
- Asymmetric Cross-Guided Attention Network for Actor and Action Video Segmentation From Natural Language Query (ICCV 2019 · similarity 0.91)
- End-to-End Joint Semantic Segmentation of Actors and Actions in Video (ECCV 2018 · similarity 0.90)
- Collaborative Spatial-Temporal Modeling for Language-Queried Video Actor Segmentation (CVPR 2021 · similarity 0.90)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.87)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.87)
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.
