- The main project objective is to develop a iOS mobile app that helps visually impaired individuals navigate safely and independently, especially when crossing roads.
- The app will use AI image recognition technology to identify safe crossing areas and obstacles.
- It will provide real-time information about the surroundings and enhance the user experience.
- Interviews were conducted with visually impaired individuals, including those with partial vision and full blindness.
- Different types of visual impairments and their specific needs were identified.
- Challenges include difficulties in transportation, identifying objects, and using technology effectively.
- Instance Segmentation (Ultralytics YOLOv8l-seg)
- OpenCV for crosswalk angle detection and obstacle detection
- Voice prompts, with adjustable voice speed in real-time.
- GitHub open data (Crosswalks-Detection-using-YOLO) / Self-labeling / 713 images
- All training data opensource in Roboflow Open datasets
- Self-captured photos and videos (designed for different daytime, nighttime, crowd, weather, and regional variations) / Self-labeling
| Box | Mask | |
|---|---|---|
| Precision | 99.9% | 97.4% |
| Recall | 99.3% | 99.3% |
| mAP50 | 99.5% | 99.5% |
| mAP50-95 | 97.4% | 96.1% |
Create a clean environment
conda create --name cleanenv python=3.8.1
conda activate cleanenv
Install dependencies
pip install ultralytics aiohttp
or install from requirements.txt
pip install -r requirements.txt
start python server
python server.py
open the address http://localhost:8080/
This work is under GNU GENERAL PUBLIC LICENSE, check LICENSE file for details. All rights reserved to Ultralytics for the YOLO model.