Research map / Label-efficient and robust learning / Recognition and long-tail
Differentiable Zooming for Multiple Instance Learning on Whole-Slide Images
Kevin Thandiackal, Boqi Chen, Pushpak Pati, Guillaume Jaume, Drew F. K. Williamson, Maria Gabrani, Orcun Goksel
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 differentiable Top-K instead of random selection at lowest magnification (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 61.0 -> 68.1 | Improved |
| use differentiable Top-K instead of random 4K selection at 2.5x (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 59.6 -> 68.1 | Improved |
| use differentiable Top-K instead of non-differentiable Top-K (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 59.9 -> 68.1 | Improved |
| use dual gated attention instead of single gated attention (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 59.6 -> 68.1 | Improved |
| add lowest-magnification features to highest-magnification features (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 62.7 -> 68.1 | Improved |
| use sum-pooling instead of concatenation of magnification features (ablation) | ZoomMIL-Eff · BRIGHT (ZoomMIL-Eff) | weighted F1: 64.9 -> 68.1 | Improved |
| add third magnification | ZoomMIL · CRC | weighted F1: 90.3 -> 92.0 | Improved |
| add third magnification | ZoomMIL · BRIGHT | weighted F1: 66.0 -> 68.3 | Improved |
| ZoomMIL vs CLAM-SB | ZoomMIL · BRIGHT | weighted F1: 63.1 -> 68.3 | Improved |
| ZoomMIL vs TransMIL with lowest magnification 10x | ZoomMIL · CAMELYON16 | accuracy: 85.3 -> 84.2 | Hurt |
| zooming instead of single-magnification CLAM-SB | ZoomMIL · BRIGHT | TFLOPs: 16.45 -> 1.29 | Improved |
| large vs small metastatic regions | ZoomMIL · CAMELYON16 validation | weighted F1: 86.7 -> 96.2 | Mixed |
Closest papers by meaning
- DGR-MIL: Exploring Diverse Global Representation in Multiple Instance Learning for Whole Slide Image Classification (ECCV 2024 · similarity 0.90)
- CaMIL: Causal Multiple Instance Learning for Whole Slide Image Classification (AAAI 2024 · similarity 0.90)
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.89)
- Fast and Accurate Gigapixel Pathological Image Classification with Hierarchical Distillation Multi-Instance Learning (CVPR 2025 · similarity 0.89)
- Dynamic Policy-Driven Adaptive Multi-Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.88)
- Queryable Prototype Multiple Instance Learning with Vision-Language Models for Incremental Whole Slide Image Classification (AAAI 2025 · similarity 0.88)
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.
