This project is a computer vision application for detecting "No Entry" signs in images. It utilizes a combination of three different techniques to achieve this:

• Viola-Jones Algorithm: A robust object detection method that uses a cascade of classifiers trained on positive and negative image samples.
• Hough Transform for Circles: A feature extraction technique used to identify circular shapes within an image.
• Template Matching: A method for finding small parts of an image that match a template image.

The system can be run on a single image or a directory of images. It evaluates its performance using the True Positive Rate (TPR) and F1 Score metrics.

The shape_detection.py script is the main entry point of the application. It performs the following steps:

1. Image Loading: Loads the input image(s) from the specified path or the ./No_entry/ directory.
2. Viola-Jones Detection: Applies the pre-trained Viola-Jones classifier (NoEntrycascade/cascade.xml) to detect potential sign regions.
3. Hough Circle Transform: Detects circles in the image using the Hough Transform.
4. Template Matching: Matches a template of the "No Entry" sign (no_entry.jpg) against the image.
5. Candidate Fusion: Combines the results from the different detectors and uses non-maximum suppression to eliminate duplicate detections.
6. Colour Filtering: Filters the detected regions based on the presence of red, a dominant colour in "No Entry" signs.
7. Evaluation: Compares the final detections against the ground truth data in groundtruth.txt and calculates the TPR and F1 Score.
8. Output: Saves the images with the detected signs in the ./Output/ directory.

• shape_detection.py: The main script that orchestrates the detection process.
• violaJones.py: Implements the Viola-Jones object detection.
• houghSpace.py: Implements the Hough Transform for circle detection.
• centralController.py: A helper class for calculating evaluation metrics like IoU, TPR, and F1-score.
• groundtruth.txt: Contains the ground truth bounding boxes for the images in the No_entry directory.
• NoEntrycascade/cascade.xml: The pre-trained cascade classifier for the Viola-Jones algorithm.
• no_entry.jpg: The template image used for template matching.
• negatives/: A directory containing negative samples used for training the classifier.
• No_entry/: A directory containing the test images with "No Entry" signs.
• Output/: The directory where the output images with detected signs are saved.

To run the detection on the default set of images in the ./No_entry/ directory, simply execute the shape_detection.py script:

python shape_detection.py

To run the detection on a specific image, use the --image argument:

python shape_detection.py --image /path/to/your/image.jpg

This project requires the following Python libraries:

• OpenCV: pip install opencv-python
• NumPy: pip install numpy
• Matplotlib: pip install matplotlib