Cap_Live_Detection

A Python application (using tkinter) that detects and highlights a man wearing a cap from a live video feed using OpenCV and MobileNet SSD model (for person detection) . It is done using basic heuristics which checks for changes in colour, edges, or patterns in the head region.

Dataset

• Cap images: Kaggle dataset ([(https://www.kaggle.com/datasets/shivanandverma/cap-dataset)])
• Pre-trained MobileNet SSD model: OpenCV

Heuristic Approach:

1. Load pre-trained MobileNet SSD model for person detection.
2. Load cap images from Kaggle dataset.
3. Resize cap images to 100x100 pixels.
4. Define a video stream function to capture frames from webcam.
5. For each frame:
   • Detect persons using MobileNet SSD model.
   • Extract head region from detected persons.
   • Resize head region to match cap image size.
   • Use template matching (TM_CCOEFF_NORMED) to detect cap.
   • Draw bounding box and label "Cap Detected" if cap is detected.
6. Display video stream using Tkinter.

Requirements

• OpenCV 4.x
• TensorFlow 2.x
• Tkinter
• Python 3.x

Future Work

• Improve cap detection accuracy.
• Expand dataset for diverse cap styles.
• Integrate with other computer vision applications.

Sample