Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Generalized Few-Shot Video Classification With Video Retrieval and Feature Generation
Yongqin Xian, Bruno Korbar, Matthijs Douze, Lorenzo Torresani, Bernt Schiele, Zeynep Akata
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 |
|---|---|---|---|
| replace 2D ResNet features with fine-tuned R(2+1)D two-stage learning (TSL) | few-shot video classification · Kinetics 5-way 1-shot | top-1 accuracy: 73.0 -> 92.5 | Improved |
| replace 2D ResNet features with R(2+1)D TSL | TAM · SomethingV2 5-way 5-shot | top-1 accuracy: 52.3 -> 80.1 | Improved |
| add Sports1M pretraining on top of base-class representation learning (ablation) | TSL · Kinetics100 5-way 1-shot | top-1 accuracy: 48.9 -> 92.5 | Improved |
| add base-class representation learning on top of pretraining (ablation) | TSL · Kinetics100 5-way 1-shot | top-1 accuracy: 27.1 -> 92.5 | Improved |
| add tag-retrieved videos without denoising or clip selection (ablation) | TSL · Kinetics100 5-way 1-shot | top-1 accuracy: 92.5 -> 91.4 | Hurt |
| add batch denoising to retrieval (ablation) | TSL w/ Retrieval · Kinetics100 5-way 1-shot | top-1 accuracy: +0.7 | Improved |
| add best clip selection to retrieval (ablation) | TSL w/ Retrieval · Kinetics100 5-way 1-shot | top-1 accuracy: +2.8 | Improved |
| add tag-retrieved videos (full retrieval) | TSL · Kinetics 5-way 1-shot | top-1 accuracy: 92.5 -> 95.3 | Improved |
| add VFGAN generated features | TSL · SomethingV2 5-way 1-shot | top-1 accuracy: 59.1 -> 66.7 | Improved |
| add retrieval | TSL · Kinetics 5-way 5-shot | top-1 accuracy: 97.8 -> 97.8 | No clear effect |
| add VFGAN features | TSL · SomethingV2 GFSV 5-way 1-shot | harmonic mean: 0.0 -> 30.2 | Improved |
| add retrieval | TSL · Kinetics GFSV 5-way 1-shot | harmonic mean: 13.8 -> 23.7 | Improved |
Closest papers by meaning
- Rethinking Generalization in Few-Shot Classification (NeurIPS 2022 · similarity 0.90)
- A Closer Look at Few-shot Classification (ICLR 2019 · similarity 0.90)
- LibFewShot: A Comprehensive Library for Few-Shot Learning (TPAMI 2023 · similarity 0.90)
- A Baseline for Few-Shot Image Classification (ICLR 2020 · similarity 0.90)
- Boosting Few-Shot Visual Learning With Self-Supervision (ICCV 2019 · similarity 0.89)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · 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.
