COLORA: Efficient Fine-Tuning for Convolutional Models with a Study Case on Optical Coherence Tomography Image Classification

TL;DR

CoLoRA is a parameter-efficient fine-tuning method for CNNs that reduces training parameters by 80%, maintains inference complexity, and improves accuracy on OCT image classification tasks.

Contribution

The paper introduces CoLoRA, a novel low-rank adaptation technique for CNNs that significantly reduces trainable parameters while enhancing performance on medical imaging tasks.

Findings

Achieves up to 1% accuracy and 0.013 AUC improvements.

Reduces per-epoch training time by nearly 20%.

Maintains original model size and inference complexity.

Abstract

We introduce CoLoRA (Convolutional Low-Rank Adaptation), a parameter-efficient fine-tuning method for convolutional neural networks (CNNs). CoLoRA extends LoRA to convolutional layers by decomposing kernel updates into lightweight depthwise and pointwise components.This design reduces the number of trainable parameters to 0.2 compared to conventional fine-tuning, preserves the original model size, and allows merging updates into the pretrained weights after each epoch, keeping inference complexity unchanged. On OCTMNISTv2, CoLoRA applied to VGG16 and ResNet50 achieves up to 1 percent accuracy and 0.013 AUC improvements over strong baselines (Vision Transformers, state-space, and Kolmogorov Arnold models) while reducing per-epoch training time by nearly 20 percent. Results indicate that CoLoRA provides a stable and effective alternative to full fine-tuning for medical image classification.