https://github.com/colasgael/photo2drawing

Automatic generation of pencil drawings from photos
https://github.com/colasgael/photo2drawing

clustering edge-detection image-processing image-to-image-translation

Last synced: 5 days ago
JSON representation

Automatic generation of pencil drawings from photos

• Host: GitHub
• URL: https://github.com/colasgael/photo2drawing
• Owner: ColasGael
• License: mit
• Created: 2019-02-12T04:07:20.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-06-11T21:11:34.000Z (over 5 years ago)
• Last Synced: 2024-06-11T17:39:53.481Z (5 months ago)
• Topics: clustering, edge-detection, image-processing, image-to-image-translation
• Language: Jupyter Notebook
• Size: 128 MB
• Stars: 4
• Watchers: 3
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# photo2drawing

by Gael Colas, Hubert Teo and Andrew Deng, graduate students from Stanford University.

This is our final project for the EE368: "Digital Image Processing" class at Stanford (2019).

Goal : Automatic generation of pencil color drawings from color photos.

To achieve this image translation task with build a pipeline. The pipeline is composed of 3 components: get a line sketch, get a colorization and blend/ apply the colors over the line sketch.

To get more details about our 'photo2drawing' pipeline, please refer to our final report 'ee368_final-report' located at the root.

## Code structure

Folder 'saliency-mask': experiment on using saliency maps to select regions of interest.

Folder 'line-integral-convolution': Python code to compute Line Integral Convolution.

 - line_integral_convolution.m: Matlab implementation of LIC

 - lic.py: Python/OpenCV/Scipy implementation of LIC which is much faster than Matlab

Folder 'baseline': gathers code for our 'baseline' pipeline.

 - line sketch generation: Canny Edge Detector

 - color generation: Clustering in the LAB-space

 

Folder 'segment': modification of our baseline to select colors based on Region Segmentation.

 - line sketch generation: Canny Edge Detector

 - color generation: mean colors of extracted regions (Region Segmentation)

Folder 'final-model': final 'photo2drawing' pipeline.

 - line sketch generation: Line Integral COnvolution

 - color generation: mean colors of extracted regions (Region Segmentation)

For these 3 pipelines, you can find the core code (.m files). 

We also provided interactive notebooks that shows intermediate outputs of the pipelines.

 

Folder 'images': gathers our different method's output result on our test set

 - subfolder 'raw': our test set images (sample data)

 - subfolder 'baseline': outputs of the 'baseline' pipeline. 

 - subfolder 'segment': outputs of the 'segment' pipeline.

 - subfolder 'saliency-map': saliency map outputs on the test images.

 - subfolder 'lic': line integral convolution on the test images.

 - subfolder 'final-model': outputs of the 'final-model' pipeline.

 - subfolder 'ps_': outputs of applying Adobe Photoshop filter  on the test images. (To allow for quality comparison)

## Run code

For an image imported in Matlab as an RGB image: 'im\_rgb = im2double(imread(''))'

For each of the 3 pipelines (see overhead description), you can run the pipeline on it in 2 different ways:

 - apply pipeline with tuned hyperparameters

Syntax: '\_hp(im_rgb)'

 - launch a GUI for real-time customization along predefined modification axes

Syntax: 'gui_(im_path)'