65 lines
1.9 KiB
Markdown
65 lines
1.9 KiB
Markdown
# Image Processing with YOLO and Inpainting
|
|
|
|
This project provides a framework for removing people from images using YOLO for object detection and OpenCV or LaMa for inpainting. You can configure the behavior of the program by editing the `conf.py` file.
|
|
|
|
## Configuration
|
|
|
|
Edit the `conf.py` file to select the processing scheme and customize its parameters:
|
|
|
|
```python
|
|
scheme = 'A' # Choose 'A' or 'B'
|
|
|
|
schemeA = {
|
|
'image_name': 'd1.jpg', # Input image name
|
|
'is_lama': False, # Use LaMa for inpainting (True/False)
|
|
'inpaint_radius': 3, # Radius for OpenCV inpainting
|
|
'is_gaussianblur': True, # Apply Gaussian blur to the mask
|
|
'gaussian_radius': 9 # Gaussian blur radius (odd number)
|
|
}
|
|
|
|
schemeB = {
|
|
'image_name': 'd1.jpg', # Input image name
|
|
'gaussian_radius': 51, # Gaussian blur radius (odd number)
|
|
'is_use_fill_color': False, # Use a fill color for blending
|
|
'fill_color': [227, 234, 244] # RGB fill color
|
|
}
|
|
```
|
|
|
|
- **Scheme A**: Uses OpenCV or LaMa for inpainting.
|
|
- **Scheme B**: Uses OpenCV for blending without inpainting.
|
|
|
|
## How to Run
|
|
|
|
1. Install the required dependencies:
|
|
```bash
|
|
pip install -r requirements.txt
|
|
```
|
|
|
|
2. Configure the `conf.py` file to select the desired scheme and parameters.
|
|
|
|
3. Run the main script:
|
|
```bash
|
|
python main.py
|
|
```
|
|
|
|
## Outputs
|
|
|
|
Processed images will be saved in the `outputs/` directory with filenames indicating the scheme and input image.
|
|
|
|
## Directory Structure
|
|
|
|
```
|
|
├── images/ # Input images
|
|
├── outputs/ # Processed images
|
|
├── conf.py # Configuration file
|
|
├── main.py # Main script
|
|
├── schemeA.py # Scheme A implementation
|
|
├── schemeB.py # Scheme B implementation
|
|
├── requirements.txt # Python dependencies
|
|
```
|
|
|
|
## Contact
|
|
|
|
If you have suggestions for improving this project or new schemes to add, feel free to contact me:
|
|
|
|
**Email**: xiadongliang88@163.com |