DINOv2 with lightweight adaptation improves trachoma image classification accuracy
Comparative Performance and Parameter-Efficient Adaptation of DINOv2 for Active Trachoma Classification
Computer Vision and Pattern Recognition
Summary
Detecting signs of trachoma, an eye disease, from photos can be tricky and expensive. This paper studies how well different computer vision models work for telling diseased eyes apart from healthy ones, using a set of eye images. The authors found that DINOv2, a pre-trained visual model, combined with a very small extra tuning step called Efficient Channel Attention, gave good accuracy with fewer extra parameters. This approach strikes a balance between performance and simplicity, making it a good benchmark for automating trachoma diagnosis from images.
What this means in practice
- •For public health teams: Use the efficient DINOv2-based classifier to support large-scale automated screening of trachoma prevalence from eye images.
- •For medical imaging developers: Implement lightweight adaptation methods like Efficient Channel Attention on pretrained models for accurate eye disease detection with minimal retraining.
Tested on one dataset.
Authors
Kibrom Gebremedhin, Hadush Hailu, Bruk Gebregziabher, Yordanos Hailu
Abstract
Automated grading of conjunctival photographs could reduce the cost and variability of trachoma prevalence surveys, but the relative value of modern pretrained visual representations, lightweight feature adaptation, and training-objective design has not been established under a common protocol. This study presents a controlled evaluation for binary classification of Trachomatous Inflammation-Follicular (TF) versus Normal using 1,546 images from the public UCSF/Lietman collection. Images are processed using the OPTED pipeline for zero-shot tarsal-conjunctiva segmentation, alignment, cropping, and standardization. We first compare six pretrained backbones using a common classification pipeline and then evaluate four lightweight adaptation mechanisms on DINOv2 ViT-B/14. Under stratified five-fold cross-validation, DINOv2 with Efficient Channel Attention (ECA) and focal-plus-center loss achieved 91.66 +/- 0.97% accuracy, 90.69 +/- 1.10% macro-F1, and 96.06 +/- 0.71% AUC. ECA introduces only five learnable parameters while matching the performance of substantially larger alternatives. Objective ablation further showed that ECA did not consistently improve plain DINOv2 across loss functions; the lowest-variance 91.66% accuracy was obtained with cross-entropy plus center loss. Overall, the fine-tuned DINOv2 representation provided most of the predictive performance, while ECA offered a highly parameter-efficient refinement whose effect depended on the training objective. The resulting workflow provides a reproducible benchmark for active trachoma image classification.