22 lines
1.1 KiB
Markdown
22 lines
1.1 KiB
Markdown
# GSOC-25: Traffic Intersection Monitoring with OpenVINO
|
|
|
|
This project develops a real-time system to detect traffic objects at intersections. It uses YOLOv11 and YOLOv12 deep learning models optimized with OpenVINO to identify vehicles, pedestrians, and traffic signs efficiently on Intel hardware.
|
|
|
|
## Current Progress (Week 1)
|
|
|
|
- Built the main detection pipeline
|
|
- Tested different YOLO models for accuracy and speed
|
|
- Created vehicle classification based on size and shape
|
|
- Developed image processing and visualization tools
|
|
- Added tracking to maintain object consistency between frames
|
|
- Implemented filtering to remove false positives and overlapping detections
|
|
|
|
## FeaturesD:\Downloads\finale6\khatam\qt_app_pyside
|
|
|
|
- Train custom YOLOv12n models using traffic data from the COCO dataset
|
|
- Convert models from PyTorch format to OpenVINO IR format
|
|
- Quantize models to INT8 for faster inference without losing accuracy
|
|
- Run detection on images, video files, and webcam streams
|
|
- Detect common traffic classes such as cars, trucks, pedestrians, and traffic lights
|
|
- Deploy models on CPU, GPU, and other OpenVINO-supported devices
|