Research map / Video / Video segmentation and tracking
Rethinking Zero-Shot Video Classification: End-to-End Training for Realistic Applications
Biagio Brattoli, Joseph Tighe, Fedor Zhdanov, Pietro Perona, Krzysztof Chalupka
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 |
|---|---|---|---|
| Train the visual encoder end-to-end rather than keeping it fixed and pretrained (ablation) | Video ZSL model · UCF101 | Top-1 accuracy: 26.8 -> 43.0 (50 classes); 19.8 -> 35.1 (101 classes) | Improved |
| Add SUN synthetic-video pretraining (ablation) | End-to-end video ZSL model · UCF101 | Top-1 accuracy: 43.0 -> 45.6 (50 classes); 35.1 -> 36.8 (101 classes) | Improved |
| Average visual embeddings from multiple test-time snippets (ablation) | End-to-end video ZSL model · UCF101 | Top-1 accuracy: 43.0 -> 48.0 (50 classes); 35.1 -> 37.6 (101 classes) | Improved |
| Train with more training classes rather than concentrating data in fewer classes (ablation) | End-to-end video ZSL model · UCF101 and HMDB51 | accuracy | Improved |
| Choose semantically diverse training classes (ablation) | Video ZSL model · UCF101 test classes | test error | Improved |
| Pretrain on synthetic videos generated from SUN still images (ablation) | Video ZSL model · UCF101 | test error: up to 10% points | Improved |
| Pretrain on synthetic SUN videos (ablation) | Video ZSL model under large train-test class distance · UCF101 | accuracy | Improved |
| Use a trainable end-to-end visual encoder instead of fixed pretrained features (ablation) | Video ZSL model · UCF101 | accuracy under large domain shift | Improved |
Closest papers by meaning
- Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective (ICCV 2019 · similarity 0.93)
- Learning to Model Relationships for Zero-Shot Video Classification (TPAMI 2020 · similarity 0.92)
- Multi-Task Zero-Shot Action Recognition with Prioritised Data Augmentation (ECCV 2016 · similarity 0.90)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.90)
- On Zero-Shot Recognition of Generic Objects (CVPR 2019 · similarity 0.90)
- Generalized Few-Shot Video Classification With Video Retrieval and Feature Generation (TPAMI 2021 · similarity 0.89)
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.
