FaceDetectCV

A lightweight OpenCV project for real-time face detection in both static images and webcam video streams using a Haar Cascade classifier.

Project Overview

FaceDetectCV demonstrates a classic computer vision workflow:

• Load a pre-trained Haar Cascade model (haarcascade_frontalface_default.xml)
• Convert frames/images to grayscale
• Detect faces with detectMultiScale(...)
• Draw bounding boxes around detected faces
• Visualize results in an OpenCV window

This is a simple, beginner-friendly starter repository for developers who want to understand face detection basics before moving on to deep learning-based approaches.

Features

• Face detection from a single image
• Face detection from live webcam feed
• Uses OpenCV's built-in Haar Cascade model
• Minimal dependencies and easy setup

Repository Structure

.
├── detect_face_image.py                # Detect faces in an image
├── detect_face_video.py                # Detect faces in webcam stream
├── haarcascade_frontalface_default.xml # Pre-trained Haar Cascade model
├── requirements.txt                    # Python dependencies
├── fac_recog.jpg                       # Example image asset
├── pic2.jpeg                           # Example image asset
└── README.md

Requirements

• Python 3.8+
• Webcam (for detect_face_video.py)
• OS with GUI support for OpenCV windows (cv2.imshow)

Install dependencies:

pip install -r requirements.txt

Note: requirements.txt currently pins older OpenCV/Numpy versions. If installation fails on modern Python versions, consider upgrading those pins.

Usage

1) Detect faces in an image

python detect_face_image.py

By default, detect_face_image.py reads a hardcoded image file:

img = cv2.imread('maaya_img.jpg')

Update this path to one of your own images (or an existing file in this repo) before running.

2) Detect faces in live webcam video

python detect_face_video.py

Controls:

• Press Esc to close the video window and stop the program.

How It Works

Both scripts follow the same detection pipeline:

1. Initialize the Haar Cascade classifier
2. Read image/frame input
3. Convert input to grayscale
4. Run face_cascade.detectMultiScale(gray, 1.1, 4)
5. Draw rectangles around detections
6. Display output

Limitations

• Haar Cascades are fast but less robust than modern deep learning detectors
• Detection quality can drop in poor lighting, occlusion, or extreme face angles
• Current scripts are minimal and use hardcoded file/webcam inputs

Future Scope:

• Add CLI arguments for input image/video path
• Save output image/video to disk
• Add confidence filtering and configurable parameters
• Add support for processing video files
• Upgrade dependencies for modern Python compatibility
• Add unit/integration tests