Research map / Vision-language and multimodal / CLIP and zero-shot
Cross-Modal Representation Learning for Zero-Shot Action Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 train visual classification and masked language modeling, then transfer semantic relatedness into visual prototypes (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 48.9 -> 54.7 | Helped |
| add masked language modeling to the classification task (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 47.4 -> 48.9 | Helped |
| use cross-modal attention in the classification task instead of visual-only attention (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 54.7 -> 39.0 | Hurt |
| replace the Transformer with an RNN (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 54.7 -> 28.2 | Hurt |
| replace the Transformer with an LSTM (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 54.7 -> 29.7 | Hurt |
| use R(2+1)D as the feature encoder instead of a 2D backbone (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 52.5 -> 59.5 | Helped |
| reduce inference from 25 clips to 1 clip (ablation) | ResT 18 · UCF101 0/50 | top-1 accuracy (%): 54.7 -> 54.0 | No clear effect |
| use ResNet-101 with 4 frames rather than ResNet-18 with 16 frames (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 54.7 -> 57.0 | Helped |
| classify in visual space using semantic transfer rather than map into semantic space (ablation) | ResT · UCF101 0/50 | top-1 accuracy (%): 48.9 -> 54.7 | Helped |
| fine-tune pretrained C3D features end-to-end (ablation) | ResT with C3D · UCF101 | top-1 accuracy (%): 40.5 -> 52.7 | Helped |
| use ResNet-101 instead of ResNet-18 (ablation) | ResT · UCF101 50% classes | top-1 accuracy (%): 54.7 ± 2.3 -> 58.7 ± 3.3 | Helped |
| use ResNet-101 instead of ResNet-18 (ablation) | ResT · HMDB51 50% classes | top-1 accuracy (%): 39.3 ± 3.5 -> 41.1 ± 3.7 | Helped |
Closest papers by meaning
- ZeroMamba: Exploring Visual State Space Model for Zero-Shot Learning (AAAI 2025 · similarity 0.89)
- Semantic-guided Cross-Modal Prompt Learning for Skeleton-based Zero-shot Action Recognition (CVPR 2025 · similarity 0.89)
- Beyond Image Classification: A Video Benchmark and Dual-Branch Hybrid Discrimination Framework for Compositional Zero-Shot Learning (CVPR 2025 · similarity 0.89)
- Progressive Semantic-Guided Vision Transformer for Zero-Shot Learning (CVPR 2024 · similarity 0.88)
- Language-Driven Cross-Modal Classifier for Zero-Shot Multi-Label Image Recognition (ICML 2024 · similarity 0.88)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 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.