Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Delta-encoder: an effective sample synthesis method for few-shot object recognition
Eli Schwartz, Leonid Karlinsky, Joseph Shtok, Sivan Harary, Mattias Marder, Abhishek Kumar, Rogerio Feris, Raja Giryes, Alex Bronstein
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 |
|---|---|---|---|
| Use learned nonlinear feature deltas instead of linear offsets (ablation) | Delta-encoder · AWA2, APY, SUN, CUB (one-shot average) | accuracy: 75.2 -> 84.3 | Improved |
| Condition synthesis on a class example instead of class attributes (ablation) | Delta-encoder · AWA2, APY, SUN, CUB (one-shot average) | accuracy: 63.4 -> 81.4 | Improved |
| Sample Z from the training set rather than random noise (ablation) | Class-example-conditioned synthesis · AWA2, APY, SUN, CUB (one-shot average) | accuracy: 81.4 -> 81.9 | Improved |
| Remove input denoising (ablation) | Class-example-conditioned synthesis · AWA2, APY, SUN, CUB (one-shot average) | accuracy: 81.9 -> 81.9 | No clear effect |
| Add the anchor example Y as an encoder input (ablation) | Autoencoder with training-set Z sampling · AWA2, APY, SUN, CUB (one-shot average) | accuracy: 81.9 -> 84.3 | Improved |
| Use the full Delta-encoder rather than the nearest-neighbor baseline | Few-shot object recognition · Caltech-256, 1-shot 5-way | accuracy: 51.3 -> 73.2 | Improved |
| Use ImageNet-pretrained features rather than features trained on the target dataset's seen categories | Delta-encoder on CUB · CUB, one-shot | accuracy: 69.8 -> 82.2 | Improved |
Later papers that compared with this method
6 later accepted papers report Delta Encoder as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- BiCrossNet with Decoupled Dual Generators: A Parameter‑Efficient and Generalizable Few‑Shot Custom Gesture Recognition Framework (ICML 2026 · better on 1 of 1 shared result)
- Variational Feature Disentangling for Fine-Grained Few-Shot Classification (ICCV 2021 · better on 6 of 6 shared results)
- Diversity Transfer Network for Few-Shot Learning (AAAI 2020 · better on 4 of 4 shared results)
- One-Shot Image Classification by Learning to Restore Prototypes (AAAI 2020 · better on 2 of 3 shared results)
- Adversarial Feature Hallucination Networks for Few-Shot Learning (CVPR 2020 · better on 2 of 2 shared results)
- Image Deformation Meta-Networks for One-Shot Learning (CVPR 2019 · better on 2 of 2 shared results)
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
- Learning Adaptive Classifiers Synthesis for Generalized Few-Shot Learning (IJCV 2021 · similarity 0.87)
- Few-Shot Image Recognition by Predicting Parameters From Activations (CVPR 2018 · similarity 0.87)
- Dynamic Few-Shot Visual Learning Without Forgetting (CVPR 2018 · similarity 0.87)
- Meta-Learning to Detect Rare Objects (ICCV 2019 · similarity 0.86)
- Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (CVPR 2020 · similarity 0.86)
- Pushing the Limits of Simple Pipelines for Few-Shot Learning: External Data and Fine-Tuning Make a Difference (CVPR 2022 · 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.
