/Face_swapping

Primary LanguagePython

Face Swap using Stable Diffusion Inpainting

This project implements a face swap technique using a Stable Diffusion Inpainting model. The workflow includes loading images, detecting faces, creating masks, and performing inpainting to blend the detected face from a source image into a target image.

Table of Contents

Overview

The project swaps a detected face from a source image onto a target image using the Stable Diffusion Inpainting model. It uses OpenCV for face detection and mask creation, and the diffusers library for inpainting.

Dependencies

  • Python 3.7+
  • PyTorch
  • diffusers
  • OpenCV
  • Pillow (PIL)
  • numpy

Install dependencies with:

pip install torch diffusers opencv-python-headless pillow numpy

File Structure

Face_Swap/
│
├── face_swap.py               # Main script
├── image1.jpg                 # Target image
├── image2.jpg                 # Source image
└── output.jpg           # Output image

Functions

  • load_model(): Loads the pre-trained Stable Diffusion Inpainting model and moves it to GPU if available.

  • load_image(image_path): Loads an image from the specified file path using OpenCV.

  • detect_face_and_create_mask(image): Detects faces in the input image and creates a mask of the detected face region.

  • convert_to_pil(image): Converts an OpenCV image to a PIL image for compatibility with the inpainting model.

  • inpaint_face(pipe, prompt, target_image, mask_image): Uses the Stable Diffusion Inpainting model to blend the source face into the target image.

  • save_image(image, path): Saves the inpainted image to a specified path and optionally displays it.

  • main():

    1. Loads the Stable Diffusion model.
    2. Loads the source and target images.
    3. Detects the face in the source image and creates a corresponding mask.
    4. Converts the target image and mask to PIL format.
    5. Inpaints the face onto the target image.
    6. Saves and displays the resulting image.

Usage

Run the script to perform the face swap:

python face_swap.py

Make sure the paths to the source and target images are correctly specified in the script. The resulting image will be saved as output.jpg in the project directory.

Acknowledgements