Research map / Label-efficient and robust learning / Recognition and long-tail
Fourier Self-Supervision for Fine-Grained Generalized Category Discovery
Sarah Rastegar, Mina Ghadimi Atigh, Pascal Mettes, Yuki Asano, Cees Snoek
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 low-frequency contrastive loss (ablation) | SelEx (last 3 blocks) · CUB-200 (DINOv1) | accuracy (All): 76.4 -> 77.5 | Improved |
| add high-frequency contrastive loss (ablation) | SelEx (last 3 blocks) · CUB-200 (DINOv1) | accuracy (All): 76.4 -> 78.8 | Improved |
| add frequency classification loss (ablation) | SelEx (last 3 blocks) · CUB-200 (DINOv1) | accuracy (All): 76.4 -> 78.0 | Improved |
| combine low and high contrastive losses (ablation) | SelEx (last 3 blocks) · CUB-200 (DINOv1) | accuracy (All): 76.4 -> 80.0 | Improved |
| add classification loss on top of both contrastive losses (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 80.0 -> 80.2 | Improved |
| add classification loss alone hurts known classes (ablation) | SelEx (last 3 blocks) · CUB-200 (DINOv1) | accuracy (Known): 72.4 -> 71.2 | Hurt |
| increase alpha_low to 1 from 0.1 (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 80.2 -> 69.2 | Hurt |
| use alpha_low 0.1 versus 0 (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 77.9 -> 80.2 | Improved |
| use alpha_high 0.1 versus 0 (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 78.1 -> 80.2 | Improved |
| use alpha_cls 0.1 versus 0 (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 79.7 -> 80.2 | Improved |
| raise SNR cutoff from 15 dB to 25 dB (ablation) | FourEx · CUB-200 (DINOv1) | accuracy (All): 80.2 -> 77.1 | Hurt |
| apply Fourier self-supervision to SelEx with DINOv1 | SelEx (last 3 blocks) · Stanford-Cars (DINOv2) | accuracy (All): 82.2 -> 80.1 | Hurt |
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.
- SelEx — SelEx: Self-Expertise in Fine-Grained Generalized Category Discovery (ECCV 2024) · better on 7 of 8 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
- SpectralGCD: Spectral Concept Selection and Cross-modal Representation Learning for Generalized Category Discovery (ICLR 2026 · similarity 0.89)
- Generalized Category Discovery under Domain Shift: A Frequency Domain Perspective (NeurIPS 2025 · similarity 0.87)
- Solving the Catastrophic Forgetting Problem in Generalized Category Discovery (CVPR 2024 · similarity 0.87)
- SEAL: Semantic-Aware Hierarchical Learning for Generalized Category Discovery (NeurIPS 2025 · similarity 0.86)
- HiLo: A Learning Framework for Generalized Category Discovery Robust to Domain Shifts (ICLR 2025 · similarity 0.86)
- SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning (ICLR 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.
