How to Enlarge Your Images with AI using Real-ESRGAN

In this guide, we will go through the process of setting up the Real-ESRGAN package for image super-resolution. Real-ESRGAN allows you to enhance the quality of images using deep learning techniques. Follow these steps to clone the repository, install the necessary dependencies, and perform inference on your images.

Prerequisites

Ensure you have Python 3.10 installed on your system.
Make sure you have Git installed to clone the repository.

Step 1: Clone the Repository

Open your terminal or command prompt and run the following commands to clone the Real-ESRGAN repository:

git clone https://github.com/xinntao/Real-ESRGAN.git
cd Real-ESRGAN

Step 2: Install Dependent Packages

Real-ESRGAN relies on several packages for training and inference. Install the required dependencies by running:

# Install specific versions of NumPy
pip install numpy<2

# Install PyTorch and torchvision for CUDA 11.6
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 -f https://download.pytorch.org/whl/torch_stable.html

# Install BasicSR
pip install basicsr

# Install facexlib and gfpgan for face enhancement
pip install facexlib
pip install gfpgan

# Install other requirements
pip install -r requirements.txt

# Set up the package
python setup.py develop

Step 3: Check CUDA Version (Optional)

You can check your NVIDIA GPU status and installed CUDA version using the following command:

nvidia-smi

Note: Checking the CUDA version is optional. You only need to do this if you encounter any errors while running the following command to install PyTorch:

pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 -f https://download.pytorch.org/whl/torch_stable.html

If you face issues during installation, verifying your CUDA version can help troubleshoot the problem.

Step 4: Download Pre-trained Models

Download the pre-trained model for Real-ESRGAN using the following command or by visiting the Real-ESRGAN releases page:

wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth

Ensure that the downloaded model file is placed in the appropriate directory as specified in the inference script.

Step 5: Running Inference

You can use the inference_realesrgan.py script to enhance images. The usage format for the script is as follows:

python inference_realesrgan.py -n RealESRGAN_x4plus -i infile -o outfile [options]...

Common Command Example

To perform image enhancement using the X4 model with face enhancement, use the following command:

python inference_realesrgan.py -n RealESRGAN_x4plus -i inputs --outscale 3.5 --face_enhance

Command Options

-h: Show help information.
-i --input: Specify the input image or folder. Default is inputs.
-o --output: Specify the output folder. Default is results.
-n --model_name: Choose the model name. Default is RealESRGAN_x4plus.
-s, --outscale: Set the final upsampling scale of the image. Default is 4.
--suffix: Define the suffix for the restored image. Default is out.
-t, --tile: Set tile size for testing. Use 0 for no tiling. Default is 0.
--face_enhance: Use GFPGAN to enhance faces. Default is False.
--fp32: Use full precision during inference. Default is fp16.
--ext: Specify the image extension. Options are auto, jpg, or png. Default is auto.

Example Command to Enhance Faces

To enhance faces in images located in the inputs folder, run:

python inference_realesrgan.py -n RealESRGAN_x4plus -i inputs --face_enhance

Check Your Processed Images After executing the inference command, navigate to the results folder. Here, you will find the final output images enhanced by Real-ESRGAN. This folder will contain the processed images, allowing you to compare them with the original inputs.

Conclusion

You have successfully set up Real-ESRGAN and performed image super-resolution. By following these steps, you can enhance the quality of your images, bringing out fine details and textures. This process can be especially useful in various fields such as photography, digital art, and more.

Additional Resources

With Real-ESRGAN, you can transform low-resolution images into stunning high-resolution masterpieces! Enjoy enhancing your images!

SeeB4Coding