Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
TAFE-Net: Task-Aware Feature Embeddings for Low Shot Learning
Xin Wang, Fisher Yu, Ruth Wang, Trevor Darrell, Joseph E. Gonzalez
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 |
|---|---|---|---|
| Add the embedding loss (ablation) | TAFE-Net · CUB GZSL | harmonic mean (H): 45.4 -> 49.2 | Improved |
| Add the embedding loss (ablation) | TAFE-Net · AWA1 GZSL | harmonic mean (H): 58.8 -> 63.2 | Improved |
| Add the embedding loss (ablation) | TAFE-Net · AWA2 GZSL | harmonic mean (H): 47.2 -> 52.2 | Improved |
| Add the embedding loss (ablation) | TAFE-Net · aPY GZSL | harmonic mean (H): 30.5 -> 36.8 | Improved |
| Add the embedding loss (ablation) | TAFE-Net · SUN GZSL | harmonic mean (H): 33.1 -> 33.0 | No clear effect |
| Use one-hot task IDs instead of word2vec (ablation) | TAFE-Net with ResNet-101 features · MITStates unseen attribute-object pairs | AP: 16.2 -> 16.1 | No clear effect |
| Use one-hot task IDs instead of word2vec (ablation) | TAFE-Net with VGG-16 features · MITStates unseen attribute-object pairs | AP: 16.3 -> 16.3 | No clear effect |
| Use one-hot task IDs instead of word2vec (ablation) | TAFE-Net with VGG-19 features · MITStates unseen attribute-object pairs | AP: 15.6 -> 16.3 | No clear effect |
| Use VGG-19 rather than ResNet-101 features with word2vec task encoding (ablation) | TAFE-Net · MITStates unseen attribute-object pairs | AP: 16.2 -> 15.6 | No clear effect |
| Use task-conditioned feature embeddings rather than a shared generic embedding | TAFE-Net · MITStates unseen attribute-object pairs | AP: +5.9 | Improved |
| Use task-conditioned feature embeddings rather than a shared generic embedding | TAFE-Net · StanfordVRD unseen SPO triplets | AP: +6.5 | Improved |
| Use TAFE-Net rather than the strongest listed meta-learning baseline | ImageNet few-shot classification · ImageNet few-shot (n=1) | all top-5 accuracy: 54.4 -> 55.7 | Improved |
Later papers that compared with this method
1 later accepted paper reports TAFE-Net as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Task Aligned Generative Meta-learning for Zero-shot Learning (AAAI 2021 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (CVPR 2020 · similarity 0.88)
- Attribute Prototype Network for Any-Shot Learning (IJCV 2022 · similarity 0.87)
- Task2Vec: Task Embedding for Meta-Learning (ICCV 2019 · similarity 0.87)
- A Dual Attention Network with Semantic Embedding for Few-Shot Learning (AAAI 2019 · similarity 0.87)
- Revisiting Unsupervised Meta-Learning via the Characteristics of Few-Shot Tasks (TPAMI 2022 · similarity 0.87)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · similarity 0.86)
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.
