OpticOps is an end-to-end AI system designed to automate the extraction of digital meter readings (electricity, water, gas) from images using deep learning and OCR. The project combines a computer vision pipeline with a mobile-first prototype to streamline data collection, improve accuracy, and reduce human effort in utility metering.
Manual meter reading is still the standard in many regions, which introduces delays, errors, and high operational costs. OpticOps addresses these challenges by enabling:
- Image-based meter reading via mobile devices
- Automatic character recognition using deep learning
- Real-time, structured data extraction
- Easy integration with smart monitoring systems
Image Capture → Preprocessing → Character Recognition → Text Output
yaml Copy code
- Capture: Users take photos of digital/analog meter displays using a smartphone.
- Preprocessing: Images are resized, denoised, and enhanced for better accuracy.
- Recognition: A PyTorch-based model detects and classifies characters in sequence.
- Output: Cleaned and validated text is generated and stored.
The mobile UI allows users to log in, scan meter displays, extract readings, and save or send the results. Below are key interface screens and output samples.
- Python (Backend Model)
- PyTorch (Model training & inference)
- Albumentations (Image preprocessing)
- scikit-learn (Data processing & metrics)
- Mobile UI (Prototype in Figma/Flutter)
OpticOps/ ├── config.py # Model configuration ├── dataset.py # Data preprocessing and loaders ├── engine.py # Training & evaluation loop ├── model.py # OCR model architecture (LcdModel) ├── train.py # Entry script to train the model ├── input/ # Training input images ├── models/ # Saved model weights ├── assets/ # App screenshots (used in README) └── README.md
yaml Copy code
- Install dependencies
pip install torch torchvision albumentations scikit-learn
Prepare data
Add .jpg images to the input/ folder
Create a CSV label file in format: filename,label
Update config.py to match file paths and hyperparameters
Train the model
bash Copy code python train.py Check output
Trained models saved in models/
Best checkpoint: lcd_model_best.pth
Project Status
-Model training and prediction pipeline complete -App prototype designed and ready for development -Deployment-ready format with modular Python code -Prepared for integration with mobile or web interfaces
Conclusion
-OpticOps is a scalable and robust solution for automating meter reading using AI. It brings together the efficiency of deep learning and the convenience of mobile interfaces, helping reduce human errors and enable faster, more reliable data collection.