Research map / Label-efficient and robust learning / Recognition and long-tail
Fine-Grained Generalization via Structuralizing Concept and Feature Space into Commonality, Specificity and Confounding
Zhen Wang, Jiaojiao Zhao, Qilong Wang, Yongfeng Dong, Wenlong Yu
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 |
|---|---|---|---|
| add concept structuralization (with feature structuralization) (ablation) | CFSG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 55.65 -> 66.56 | Improved |
| add feature structuralization to backbone + CS (ablation) | CFSG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 58.75 -> 66.56 | Improved |
| add feature structuralization alone to backbone (ablation) | Backbone · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 51.65 -> 55.65 | Improved |
| add concept structuralization alone to backbone (ablation) | Backbone · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 51.65 -> 58.75 | Improved |
| use fixed manual component weights instead of learnable weights (ablation) | CFSG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 55.00 -> 66.56 | Improved |
| classify with structured concept weights instead of sub-centroids (ablation) | CFSG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 59.53 -> 66.56 | Improved |
| sub-centroid classification vs FSDG baseline | FSDG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 55.65 -> 59.53 | Improved |
| replace FSDG with CFSG on ResNet-50 dual backbone | FSDG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 55.65 -> 66.56 | Improved |
| replace FSDG with CFSG on ResNet-101 | FSDG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 57.25 -> 68.44 | Improved |
| replace FSDG with CFSG on ViT-Small | FSDG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 68.26 -> 70.31 | Improved |
| replace FSDG with CFSG on ASMLP-Small | FSDG · CUB-Paintings (avg of C->P, P->C) | accuracy (%): 58.75 -> 67.19 | Improved |
| use CFSG on VMamba instead of HSSH on Birds-31 | HSSH · Birds-31 (avg of 6 transfers) | accuracy (%): 90.69 -> 90.52 | No clear effect |
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.
- PAN (DA) — PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (CVPR 2021) · better on 3 of 3 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
- Fine-Grained Recognition: Accounting for Subtle Differences between Similar Classes (AAAI 2020 · similarity 0.89)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.88)
- From Coarse to Fine-Grained Open-Set Recognition (CVPR 2024 · similarity 0.87)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.87)
- Is Large-scale Pretraining the Secret to Good Domain Generalization? (ICLR 2025 · similarity 0.87)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 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.
