Skip to content

WEBCAM_SETUP.md

Latest commit

 

History

History
114 lines (89 loc) · 3.57 KB

WEBCAM_SETUP.md

File metadata and controls

114 lines (89 loc) · 3.57 KB
Raw

Webcam Setup Guide

Overview

The Space Station Object Detection app now includes live webcam functionality for real-time detection of:

  • 🔧 Toolbox
  • 🧯 Fire Extinguisher
  • 🫧 Oxygen Tank

Prerequisites

1. Install Dependencies

pip install streamlit-webrtc av

Or install all requirements:

pip install -r requirements.txt

2. System Requirements

  • Operating System: Windows 10/11, macOS, or Linux
  • Webcam: Any USB webcam or built-in camera
  • Browser: Chrome, Firefox, Safari, or Edge (with WebRTC support)
  • Internet: Required for WebRTC connection

Usage Instructions

1. Start the Application

streamlit run app.py

2. Load a Model

  1. In the sidebar, select a trained model from the dropdown
  2. Click "Load Model" button
  3. Wait for the success message

3. Access Live Webcam

  1. Click on the "📹 Live Webcam" tab
  2. Click "Start Webcam" to begin live detection
  3. Point your camera at objects to detect
  4. View real-time detection results and statistics

Features

Live Detection

  • Real-time object detection with bounding boxes
  • Confidence scores displayed on each detection
  • Color-coded detection boxes for each object type

Live Statistics

  • Real-time detection counts by object type
  • Interactive pie chart showing detection distribution
  • Recent detection history (last 10 detections)

Session Management

  • Download detection logs as CSV
  • Generate session reports in Markdown format
  • Track detection timestamps and confidence scores

Troubleshooting

Webcam Not Working

  1. Check Browser Permissions: Allow camera access when prompted
  2. Try Different Browser: Some browsers have better WebRTC support
  3. Check Webcam: Ensure webcam is not being used by another application
  4. Test Basic Webcam: Run test_webcam.py to verify basic functionality

Performance Issues

  1. Lower Detection FPS: Reduce the detection FPS slider in settings
  2. Close Other Applications: Free up system resources
  3. Check Model Size: Use smaller models for better performance

Connection Issues

  1. Check Internet: WebRTC requires internet connection for STUN servers
  2. Firewall: Ensure firewall allows WebRTC connections
  3. Corporate Network: Some corporate networks block WebRTC

Model Information

Supported Objects

  • Toolbox (Class 0): Various types of toolboxes
  • Fire Extinguisher (Class 2): Fire extinguishers
  • Oxygen Tank (Class 1): Oxygen tanks

Model Files

  • runs/detect/train/weights/best.pt - Best model from training run 1
  • runs/detect/train2/weights/best.pt - Best model from training run 2

Advanced Configuration

Detection Settings

  • Confidence Threshold: Adjust minimum confidence for detections (0.1-0.9)
  • Detection FPS: Control how often detection runs (1-30 FPS)
  • Show Confidence: Toggle confidence score display

Performance Optimization

  • Detection runs every 3 frames by default for performance
  • Queue system prevents memory overflow
  • Automatic cleanup of old detection history

Security Notes

  • Webcam access requires user permission
  • No video data is stored permanently
  • All processing happens locally in the browser
  • Detection results can be downloaded for analysis

Support

If you encounter issues:

  1. Check the troubleshooting section above
  2. Run the test script: streamlit run test_webcam.py
  3. Verify all dependencies are installed correctly
  4. Check browser console for error messages