Understanding Domain Adaptation Using CORAL in Computer Vision

This paper investigates whether Domain Adaptation techniques can significantly improve the performance of Convolutional Neural Networks (CNNs) in image classification across varying domains and distributions. Our work applies Deep CORAL with EfficientNetV2 for domain adaptation on the Office-31 dataset. We compare its performance to a regular EfficientNetV2 model that doesn’t use domain adaptation, measuring improvements with metrics as follows: accuracy, precision, recall, and F1 score. The CORAL-implemented model demonstrated a 4.15% average boost in average precision, recall, F1, and accuracy across all 3 trials.