SugarScope AI

Background

SugarScope AI is a machine learning application delivering real-time image classification predictions. We used Microsoft Lobe AI to train and export ONNX models, which are then deployed to infer from live RTSP camera feeds.

Features

• Real-time predictions from RTSP streaming camera input
• Intuitive GUI built with Tkinter
• Threaded frame grabbing for smooth performance

How to Use

• Install required libraries via pip install -r requirements.txt
• Update RTSP link & model path in config.xml
• Run the app: python main.py
• Use the GUI to view live inference from your RTSP stream

Optimizations

• Multithreading to accelerate camera frame capture
• Model tuning for accuracy vs. speed trade-offs
• Adjustable frame rates to match network bandwidth

Lessons Learned

• RTSP streaming requires careful buffer management
• Tkinter provides a lightweight GUI solution
• Microsoft Lobe AI simplifies model training and ONNX export

Source Code

View on GitHub