Project Overview
Objective
Built and benchmarked EV charger socket instance segmentation models for real-time use.
Stack
YOLOv8-SegYOLOv12-SegRF-DETR-SegMask R-CNN
Delivery highlights
- Built and trained instance segmentation models (YOLOv8-Seg, YOLOv12-Seg, RF-DETR-Seg, Mask R-CNN) to identify and segment multiple EV charger socket types in real time. Collected and annotated data, performed preprocessing (resize/normalize, label checks) and augmentation (flip/rotate/brightness), then trained and evaluated each model to compare accuracy and inference speed and select a practical model for use.