Unlocking the Perfect Fit: ML Approach for Apparel Size Recommendation

This study is based on the exploration of an e-commerce apparel size recommendation system through machine learning and image processing, evaluating three backbone architectures for YOLOv8—EfficientNet, ResNet, and CSPDarknet—identifying CSPDarknet as the most effective. The novelty of the model is emphasized by its conjunction with a custom-gathered dataset, further optimized by CLAHE for data augmentation and the AdamW optimizer. Validated with precise measurements on a set of numerous diverse individuals, the approach integrates C2f and SPPF blocks to enhance gradient flow and model efficiency, significantly advancing the online shopping experience by optimizing the analysis of one single user image.