Research map / Image generation / Fast diffusion sampling
Advancing Image Classification with Discrete Diffusion Classification Modeling
Omer Belhasin, Shelly Golan, Ran El-Yaniv, Michael Elad
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 |
|---|---|---|---|
| replace cross-entropy classifier with DiDiCM-CP (8 steps) | ResNet-50 (Strong Aug) · ImageNet-1k (56px, 25% data) | top-1 accuracy: 53.77 -> 59.05 | Improved |
| replace cross-entropy classifier with DiDiCM-CP (8 steps) | ResNet-50 (Weak Aug) · ImageNet-1k (56px, 25% data) | top-1 accuracy: 44.86 -> 57.92 | Improved |
| replace cross-entropy classifier with DiDiCM-CP (8 steps) | ResNet-50 (Strong Aug) · ImageNet-1k (224px, 100% data) | top-1 accuracy: 80.42 -> 80.40 | No clear effect |
| replace cross-entropy classifier with DiDiCM-CP (8 steps) | ResNet-50 (Strong Aug) · ImageNet-1k (224px, 100% data) | top-5 accuracy: 94.60 -> 95.29 | Improved |
| train DiDiCM with Weak Aug instead of standard classifier with Strong Aug | DiDiRN-50 · ImageNet-1k (56px, 25% data) | top-1 accuracy: 53.77 -> 57.92 | Improved |
| train DiDiCM with Weak Aug instead of standard classifier with Strong Aug | DiDiRN-50 · ImageNet-1k (224px, 100% data) | top-1 accuracy: 80.42 -> 77.68 | Hurt |
| increase uncertainty (lower resolution, less data) | DiDiCM vs standard classifier · ImageNet-1k (Weak Aug) | top-1 gain: gain grows to 13.1% at 56px/25% | Improved |
| use DiDiCM-CP with two diffusion steps vs more steps (ablation) | DiDiCM-CP · ImageNet-1k (56px, 100% data) | top-1 accuracy | No clear effect |
Closest papers by meaning
- Classification Diffusion Models: Revitalizing Density Ratio Estimation (NeurIPS 2024 · similarity 0.87)
- Diffusion Classifiers Understand Compositionality, but Conditions Apply (NeurIPS 2025 · similarity 0.86)
- Efficient Image-to-Image Diffusion Classifier for Adversarial Robustness (AAAI 2025 · similarity 0.86)
- An Efficient Framework for Enhancing Discriminative Models via Diffusion Techniques (AAAI 2025 · similarity 0.86)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.86)
- Elucidating the design space of classifier-guided diffusion generation (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.
