AutoReadSpeed is a Streamlit-based app that uses computer vision and OCR to extract speedometer readings from uploaded videos (digital odometers currently supported; analog odometers may be implemented in a future release). It's designed for use in automotive testing, data logging, intelligent driving analysis, and by driving enthusiasts.
- 📹 Upload a video of a speedometer
- ⏱️ Sample frames at customizable intervals (e.g., every 3 seconds)
- 🔍 On the backend, the app uses EasyOCR or Tesseract to extract speed values from each frame
- 📈 View speed trends through interactive plots
- 📤 Export results to .XLSX
git clone https://github.com/baileyarzate/AutoReadSpeed.git
cd AutoReadSpeedpython -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activatepip install -r requirements.txtstreamlit run streamlit_app_AutoReadSpeed.py📂 Outputs
- A .xlsx file with timestamps and detected speed values
- Line plots showing speed over time
🧪 Future Improvements
- Support for analog odometer detection (via dial or needle tracking)
- Visual feedback for OCR bounding boxes
- Batch processing for multiple videos
- Enhanced model filtering for noisy inputs
- Enhanced data analysis
- Optional video playback with overlayed speed values
📄 License
- This project is licensed under the MIT License.
🙋♀️ About the Author
- Created by @baileyarzate, a Data Scientist working for the U.S. Air Force as a civil servant with a passion for machine learning, vision systems, and intelligent automation.