Research map / Label-efficient and robust learning / Recognition and long-tail
Explanation-based Data Augmentation for Image Classification
Sandareka Wickramanayake, Wynne Hsu, Mong-Li Lee
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 |
|---|---|---|---|
| drop the concept-match term Delta from the utility score (ablation) | BRACE with CCNN · CUB Dense-161 | accuracy (%): 86.8 -> 85.9 | Hurt |
| drop the under-representation score beta from the utility score (ablation) | BRACE with CCNN · CUB Dense-161 | accuracy (%): 86.8 -> 86.3 | Hurt |
| drop Delta from utility (ablation) | BRACE with CCNN · CUB-Families Dense-161 | accuracy (%): 91.9 -> 91.2 | Hurt |
| drop beta from utility (ablation) | BRACE with CCNN · CUB-Families Dense-161 | accuracy (%): 91.9 -> 91.3 | Hurt |
| select added samples by utility score instead of randomly | explanation-based augmentation · CUB-Families Res-101 post-hoc | accuracy (%): 88.3 -> 91.2 | Improved |
| select samples by classifier confidence instead of utility | sample selection · CUB-Families class accuracy (Hirundinidae) | accuracy (%): 41.1 -> 32.9 | Hurt |
| augment with BRACE-selected images instead of CutMix/SnapMix mixing | CCNN ResNet-34 · CUB | accuracy (%): 80.6 -> 86.1 | Improved |
| add Metaset-based repository samples | ResNet-34 black-box · Tiny ImageNet | accuracy (%): 76.1 -> 73.1 | Hurt |
| augment with BRACE | ResNet-34 black-box · Tiny ImageNet | accuracy (%): 76.1 -> 78.3 | Improved |
| augment with BRACE for out-of-distribution generalization | ResNet-34 · ImageNet-V2-Sub | accuracy (%): 54.4 -> 70.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SnapMix — SnapMix: Semantically Proportional Mixing for Augmenting Fine-grained Data (AAAI 2021) · 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
- Interpretable Basis Decomposition for Visual Explanation (ECCV 2018 · similarity 0.90)
- CE-FAM: Concept-Based Explanation via Fusion of Activation Maps (ICCV 2025 · similarity 0.89)
- Making the Classification Explanation Faithful to the Confidence Score (CVPR 2026 · similarity 0.88)
- A Generalized Explanation Framework for Visualization of Deep Learning Model Predictions (TPAMI 2023 · similarity 0.88)
- Invertible Concept-based Explanations for CNN Models with Non-negative Concept Activation Vectors (AAAI 2021 · similarity 0.88)
- I-CEE: Tailoring Explanations of Image Classification Models to User Expertise (AAAI 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.
