https://github.com/linchuming/wssdcnn

The unofficial implementation of Weakly- and Self-Supervised Learning for Content-Aware Deep Image Retargeting via TensorFlow
https://github.com/linchuming/wssdcnn

retargeting tensorflow

Last synced: 9 months ago
JSON representation

The unofficial implementation of Weakly- and Self-Supervised Learning for Content-Aware Deep Image Retargeting via TensorFlow

• Host: GitHub
• URL: https://github.com/linchuming/wssdcnn
• Owner: linchuming
• License: mit
• Created: 2019-06-20T09:57:04.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2019-06-20T10:57:00.000Z (about 7 years ago)
• Last Synced: 2025-08-15T13:42:46.031Z (11 months ago)
• Topics: retargeting, tensorflow
• Language: Python
• Homepage:
• Size: 2.95 MB
• Stars: 8
• Watchers: 0
• Forks: 0
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# WSSDCNN

The unofficial implementation of Weakly- and Self-Supervised Learning for Content-Aware Deep Image Retargeting

### Preparation

- TensorFlow >= 1.6

- Download the param (577M) from BaiduYun: https://pan.baidu.com/s/15UvYzwriYCOiqqb6ZvhqsA (passwd: jlq5)

- Move the param files to directory `model_ckpt`

### Inference

- ` python test.py`

### Results

The approach has limits and their results are not always good.  Some good results show below (aspect ratio = 0.5) :

| Original                   | WSSDCNN                        | Bicubic                        |

| -------------------------- | ------------------------------ | ------------------------------ |

| ![](results/butterfly.png) | ![](results/butterfly_0.5.png) | ![](results/butterfly_bic.png) |

| ![](results/eagle.png)     | ![](results/eagle_0.5.png)     | ![](results/eagle_bic.png)     |

| ![](results/surfer.png)    | ![](results/surfer_0.5.png)    | ![](results/surfer_bic.png)    |