File size: 957 Bytes
5fd1c47
 
 
 
 
 
9649bde
 
5fd1c47
 
 
 
 
dcc0b46
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
---
title: EDSR Keras
emoji: 🚀
colorFrom: pink
colorTo: yellow
sdk: gradio
sdk_version: 3.18.0
python_version: 3.10.9
app_file: app.py
pinned: false
license: mit
---

This space is the demo for the EDSR (Enhanced Deep Residual Networks for Single Image Super-Resolution) model. This model surpassed the performace of the current available SOTA models.

Paper Link - https://arxiv.org/pdf/1707.02921

Keras Example link - https://keras.io/examples/vision/edsr/


TODO:

Hack to make this work for any image size. Currently the model takes input of image size 150 x 150.
We pad the input image with transparant pixels so that it is a square image, which is a multiple of 150 x 150
Then we chop the image into multiple 150 x 150 sub images
Upscale it and stich it together.

The output image might look a bit off, because each sub-image dosent have data about other sub-images.
This approach assumes that the subimage has enough data about its surroundings