Image-SuperResolution-SREZ

Image super-resolution through deep learning. This project uses deep learning to upscale 16x16 images by a 4x factor.

Dependencies:

• Python v3.6.+ (64bits)
• NumPy v1.14.+
• Pillow v5.+
• OpenCV v3.14.+ (for cv2)
• MatPlotLib v2.1.+
• MoviePy (0.2.x)
• TensorFlow v1.+

Installations

Note: For install libraries use CMD terminal.

1. Download Python 3.6.+ (last version of 64bits), and install it. Guide Video
2. Installing numpy library (if not included)

py -m pip install numpy

3. Installing OpenCV library

py -m pip install opencv-python

4. Installing Pillow library

py -m pip install Pillow

5. Installing MatPlotLib

py -m pip install matplotlib

6. Installing TensorFlow (CPU or GPU) library

#*CPU version*
py -m pip install --upgrade tensorflow

or

#*GPU version*
py -m pip install --upgrade tensorflow-gpu

7. Install Imageio library

py
import imageio
imageio.pluggins.ffmpeg.download()

For GPU Nvidia, must install CUDA v9.0 and cuDNN v9.0 (for cuDNN follow this steps)

Project

Create a folder name srez and copy the following py files and folders:

• checkpoint\ (learning process)
• dataset\ (200K celebrity faces)
• train\ (output of learning process, empty)
• srez_demo.py (Create Video animation of outputs)
• srez_input.py (dependicy of main.py)
• srez_main.py (main code)
• srez_model.py (dependicy of main.py)
• srez_train.py (For train the model with celeb faces)

Using Networks

Usage is as follows:

1. Download the Large-scale CelebFaces 1.34GB dataset file 'img_align_celeba.zip'. Unziped on dataset folder. This step may take a while as it will extract 200K images.

2. Training with default settings: py srez_main.py --run train. The script will periodically output an example batch in PNG format onto the srez/train folder, and checkpoint data will be stored in the srez/checkpoint folder.

3. After the network has trained you can also produce an animation showing the evolution of the output by running: py srez_main.py --run demo

Note: In srez_main.py file, line 64, specify the training time (default: 1200 minutes). or any other values you want to change

Reproduce: David GPU