Research map / Architectures and efficiency / Attention and transformers
A Rotation-invariant Texture ViT for Fine-Grained Recognition of Esophageal Cancer Endoscopic Ultrasound Images
Tianyi Liu, Shuaishuai S Zhuang, Jiacheng Nie, Geng Chen, Yusheng Guo, Guangquan Zhou, Jean-Louis Coatrieux, Yang Chen
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 token selection module to ViT (ablation) | ViT-B 16 · In-house EUS dataset | accuracy (%): 61.6 -> 76.3 | Improved |
| add rotation-invariant reinforcement (soft histogram) to ViT+TS (ablation) | ViT+TS · In-house EUS dataset | accuracy (%): 76.3 -> 77.4 | Improved |
| add contrastive loss to ViT+TS+RR (ablation) | SRRM-ViT · In-house EUS dataset | accuracy (%): 77.4 -> 78.7 | Improved |
| add rotation-invariant reinforcement to ViT+TS+Lcon (ablation) | ViT+TS+Lcon · In-house EUS dataset | accuracy (%): 77.9 -> 78.7 | Improved |
| add reinforcement to ViT+TS+Lcon (ablation) | SRRM-ViT · In-house EUS dataset | recall (%): 41.8 -> 41.1 | No clear effect |
| add rotation-invariant reinforcement (ablation) | ViT+TS · In-house EUS dataset | FLOPs (G): 17.0 -> 32.5 | Hurt |
| use 8 histogram bins instead of 4 (ablation) | SRRM-ViT · In-house EUS dataset | accuracy (%): 75.6 -> 78.7 | Improved |
| use 16 histogram bins instead of 8 (ablation) | SRRM-ViT · In-house EUS dataset | accuracy (%): 78.7 -> 76.2 | Hurt |
| set contrastive margin alpha to 0.7 instead of 0.3 (ablation) | SRRM-ViT · In-house EUS dataset | accuracy (%): 75.6 -> 78.7 | Improved |
| raise alpha from 0.7 to 0.9 (ablation) | SRRM-ViT · In-house EUS dataset | accuracy (%): 78.7 -> 76.5 | Hurt |
| apply SRRM-ViT instead of plain ViT on Brain Tumor MRI | ViT-B 16 · Brain Tumor MRI | accuracy (%): 96.9 -> 99.2 | Improved |
| SRRM-ViT versus CSA-MLP on Brain Tumor MRI recall | SRRM-ViT · Brain Tumor MRI | recall (%): 87.8 vs 98.5 | Hurt |
Closest papers by meaning
- Leveraging Anatomical Consistency for Multi-Object Detection in Ultrasound Images via Source-free Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.82)
- CSC-PA: Cross-image Semantic Correlation via Prototype Attentions for Single-network Semi-supervised Breast Tumor Segmentation (CVPR 2025 · similarity 0.82)
- FocusMAE: Gallbladder Cancer Detection from Ultrasound Videos with Focused Masked Autoencoders (CVPR 2024 · similarity 0.82)
- MCAT: Visual Query-Based Localization of Standard Anatomical Clips in Fetal Ultrasound Videos Using Multi-Tier Class-Aware Token Transformer (AAAI 2025 · similarity 0.82)
- Unsupervised Domain Adaptation for Anatomical Structure Detection in Ultrasound Images (ICML 2024 · similarity 0.82)
- Rotation-Agnostic Image Representation Learning for Digital Pathology (CVPR 2024 · similarity 0.81)
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.
