Project Overview
Objective
Built mobile app for EV socket detection from camera and gallery images.
Stack
React NativeExpoONNXYOLOv8
Delivery highlights
- Built a React Native (Expo) mobile app to detect and classify EV charger socket types in real time from the camera and gallery images. Trained a YOLOv8 detection model and deployed it via ONNX for on-device inference, implementing image preprocessing and an in-app overlay for bounding boxes, labels, and confidence scores.
- Supports multiple connector types: AC Type 1, AC Type 2, DC CCS Type 1, DC CCS Type 2, and CHAdeMO.
- Uses on-device ONNX inference for real-time processing without requiring backend API calls, enabling instant feedback and privacy-first operation.
- Implements image preprocessing (resize, normalization, tensorization) directly on the mobile device to meet model input requirements.
- Renders detection results with visual overlays showing bounding boxes, class labels, and confidence scores for user-friendly identification.
Back to Topic ProjectsBack to All ProjectsSystem workflow
How on-device EV socket detection works
Camera or gallery images are prepared, inferred locally with ONNX, and rendered with detection overlays without a remote inference API.Rendering workflow diagram…
Scroll horizontally to explore the full workflow on smaller screens.
System workflow
Mobile app architecture
React Native components for camera capture, image selection, and inference results display, built on top of ONNX Runtime for on-device model execution.Rendering workflow diagram…
Scroll horizontally to explore the full workflow on smaller screens.
System workflow
Supported EV charger socket types
The model detects and classifies five standardized EV connector types, enabling users to identify charging port compatibility at a glance.Rendering workflow diagram…
Scroll horizontally to explore the full workflow on smaller screens.
System workflow
On-device inference pipeline
Complete data flow from image input through model inference to detection output, all running locally on the mobile device without network requests.Rendering workflow diagram…
Scroll horizontally to explore the full workflow on smaller screens.