DESIGN DECISION
Training and deployment are separated
TensorFlow handles training and evaluation, TensorRT optimizes inference, and Android manages camera input, model selection, and overlays.
TensorFlowTensorRTSemantic SegmentationAndroid
Empowering edge perception, reshaping the mobile intelligent driving vision.
EDGE SEGMENTATION
Turning road pixels into a readable driving scene
A TensorRT-optimized segmentation model runs on Android and converts each frame into semantic regions.
PIXELper-pixel labels
GPUTensorRT acceleration
EDGEAndroid deployment
01 FRAME Camera input →
02 MODEL TensorFlow →
03 OPT TensorRT →
04 MAP Semantic mask →
05 UI Android view
TensorFlow
Camera
Segmentation
Android
CORETensorRT
CAPABILITIES
Live visual output
01Raw frame✓
02Class mask✓
03Overlay preview✓
04Model selection✓
Project Showcase
The project was successfully deployed on Android devices, achieving real-time semantic segmentation functionality.
Intelligent Driving Project Showcase Original 
Intelligent Driving Project Showcase Recognized 
Selectable Models Software Copyright Certificate
This project successfully obtained software copyright
