opencv-censor-bar

Playing around with opencv. This script detects faces and eyes in an input image using Haar cascades, then adds a censor bar over the detected eyes

Example

Input Image

Output Image

How it works

• Takes an input image file as a command-line argument
• Reads the input image using the OpenCV library
• Detects faces and eyes in the image using Haar cascades
• Converts the OpenCV image to a PIL (Python Imaging Library) image
• Adds a censor bar over the detected eyes in the image
• Saves the modified image with the censor bar as a new file named "output.jpg"

Usage

pip install opencv-python opencv-python-headless pillow

python script.py image.jpg

TODO: Improvements

• Replace Haar cascades with a more accurate model (deep learning based model? MTCNN? pretrained model from Dlib?)
• Add a confidence threshold for the detected faces and eyes to filter out false positives
• Experiment with different values for scaleFactor and minNeighbors to improve face and eye detection
• Preprocess the input image to improve robustness to different lighting conditions, rotations, and scales
• Resize very large images before processing to speed up the detection process
• Search for eyes only within the detected face regions to improve efficiency
• Add error handling for cases where the image file cannot be read, the Haar cascade files cannot be found, or other issues arise during processing