https://github.com/rsgoksel/g_esrgan

Both image and video capable ESRGAN model
https://github.com/rsgoksel/g_esrgan

esrgan gan gan-network generative-adversarial-network generator genetic-algorithm image-enhancement

Last synced: 11 days ago
JSON representation

Both image and video capable ESRGAN model

• Host: GitHub
• URL: https://github.com/rsgoksel/g_esrgan
• Owner: RsGoksel
• License: apache-2.0
• Created: 2023-06-14T21:29:39.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-09-11T17:39:26.000Z (2 months ago)
• Last Synced: 2024-09-12T02:54:53.442Z (2 months ago)
• Topics: esrgan, gan, gan-network, generative-adversarial-network, generator, genetic-algorithm, image-enhancement
• Language: Python
• Homepage:
• Size: 42 KB
• Stars: 0
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# G_ESRGAN

Both image and video capable ESRGAN model 

**Weights can be found in release** After downloading it, put it in the weights folder.

Colab Page:   [![Colab](https://img.shields.io/static/v1?label=Demo&message=Colab&color=orange)](https://colab.research.google.com/drive/1HVMSJoFMNugZMZ5mX51mraH7yf8rWPCv?authuser=2)

🚀 _**Usage of GAN model**_:

___________________________________

For the _**images**_, (PNG, JPG and any extension with cv2 library can handle)    

```bash

    python Image_GAN.py --image-path example.png

```

For the _**Videos**_, (Just mp4)

```bash

    python Video_GAN.py --video-path example.mp4

```

* Results will be in output folder after the processing

* Video results might be slow about 360p and above

_______________________________________________________________

🚀 Here are some results images and videos 

_______________________________________________________________

https://github.com/RsGoksel/G_ESRGAN/assets/80707238/7025522f-228c-4bcd-8f11-62b3c68611e1

___________________________________________________________________________

As a result of the GAN models' training dataset, inference results tend to perform better, particularly when dealing with images that include anime or other forms of artificial imagery, like CGI.

![Yugi](https://github.com/RsGoksel/G_ESRGAN/assets/80707238/b3c9bbb3-580d-4a80-93fb-581d8bd3b1b3)

_______________________________________________________________

![Lord1](https://github.com/RsGoksel/G_ESRGAN/assets/80707238/da50f7a6-ac00-4759-9b63-c2c3be29bd50)

_____________________________________________

![Lord2](https://github.com/RsGoksel/G_ESRGAN/assets/80707238/8be3898e-4f27-4ce9-bc59-dcd72daae690)

________________________________