Other Projects/Traffic Sign Recognizer
AI Traffic Sign Recognizer
A camera-based prototype that classifies local road signs using OpenCV and a convolutional neural network.
Overview
Built for Elective II coursework, this team prototype explores recognition of road signs found in Pampanga. The training pipeline crops annotated sign images, resizes them to 64 × 64 pixels, augments the training data, and trains a TensorFlow CNN. The camera script uses edge contours to find candidate regions, then displays predicted labels and confidence scores. The desktop interface adds camera selection, Start/Stop controls, adjustable confidence, and workspace screenshot saving. The reference preview shows saved classifier predictions on dataset crops; it is not a live-camera benchmark.
Features
- Train a CNN from labeled sign images and bounding-box annotations.
- Augment training images with rotations, shifts, zoom, and flips.
- Find candidate sign regions with OpenCV edge detection and contour filtering.
- Display predictions above the confidence threshold with bounding boxes and an FPS counter.
Built with
- Vision & learning
- Python, OpenCV, TensorFlow / Keras, NumPy
- Training & data
- pandas, scikit-learn, CSV annotations, joblib, Tkinter, Pillow