A lightweight and interactive object recognition system built with Streamlit, allowing users to:
- Upload two images representing different objects.
- Train a custom classifier on-the-fly using those images.
- Classify live webcam input or uploaded images into Object 1 or Object 2.
- Upload one image for each object class.
- On-the-fly model training using Logistic Regression.
- Real-time image prediction using webcam or uploaded image.
- Augments data with rotation, brightness, and contrast variations.
- Uses color, shape, texture, and SIFT-based features.
- Adjustable confidence threshold for predictions.
streamlit opencv-python numpy Pillow scikit-learn
. ├── app.py # Main Streamlit app (your uploaded code) ├── README.md # This file ├── requirements.txt # List of dependencies (optional) └── temp_images/ # Temporary folder for images (auto-created at runtime)