Research map / Video / Video segmentation and tracking
Differentiable Grammars for Videos
AJ Piergiovanni, Anelia Angelova, Michael S. Ryoo
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 differentiable grammar to I3D features | I3D · MLB-YouTube | mAP: 34.2 -> 43.4 | Improved |
| Add differentiable grammar to I3D + super-events | I3D + super-events · MLB-YouTube | mAP: 39.1 -> 44.2 | Improved |
| Add differentiable grammar to I3D features | I3D · MultiTHUMOS | mAP: 29.7 -> 32.3 | Improved |
| Add differentiable grammar to I3D + super-events | I3D + super-events · MultiTHUMOS | mAP: 36.4 -> 37.7 | Improved |
| Add differentiable grammar to I3D + TGM | I3D + TGM · MultiTHUMOS | mAP: 46.4 -> 48.2 | Improved |
| Add differentiable grammar to I3D features | I3D · Charades v1_localize | mAP: 17.2 -> 18.5 | Improved |
| Add differentiable grammar to I3D + super-events | I3D + super-events · Charades v1_localize | mAP: 19.4 -> 20.3 | Improved |
| Add differentiable grammar to I3D + TGM | I3D + TGM · Charades v1_localize | mAP: 22.3 -> 22.9 | Improved |
| Use grammar-based future generation instead of last-frame repetition | I3D video forecasting · MultiTHUMOS, 2 sec | mAP: 16.0 -> 18.6 | Improved |
| Use grammar-based future generation instead of last-frame repetition | I3D video forecasting · MultiTHUMOS, 10 sec | mAP: 12.7 -> 12.8 | Improved |
| Use grammar-based future generation instead of last-frame repetition | I3D video forecasting · MultiTHUMOS, 20 sec | mAP: 8.7 -> 10.5 | Improved |
| Use grammar-based future generation instead of last-frame repetition | I3D video forecasting · Charades, 2 sec | mAP: 13.8 -> 14.8 | Improved |
Closest papers by meaning
- Adversarial Generative Grammars for Human Activity Prediction (ECCV 2020 · similarity 0.83)
- VideoWorld: Exploring Knowledge Learning from Unlabeled Videos (CVPR 2025 · similarity 0.81)
- Diffusion On Syntax Trees For Program Synthesis (ICLR 2025 · similarity 0.81)
- Towards a theory of how the structure of language is acquired by deep neural networks (NeurIPS 2024 · similarity 0.81)
- Language Model Beats Diffusion - Tokenizer is key to visual generation (ICLR 2024 · similarity 0.81)
- Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization (ICML 2024 · similarity 0.81)
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.
