Jump to content
Martin Brooks

New method for automatic image vectorization -- the most complex SVG images on the planet!

Recommended Posts

Hello Affinity designers,

I have a lot of fun using AD to explore automatically vectorized images.

Here is a gallery: www.shapevision.gallery  

Right-click on the shape images to download, and then check them out in AD.

These svg images were produced by a new method. Although it is targeted at engineering applications, perhaps there is application for creators. Ideas?!?

You can get a sense of how the method works by trying it here: www.shapevision.technology  .

Martin Brooks

 

rasberry canes 1a.svg

Share this post


Link to post
Share on other sites

Hello Lon and VectorVonDoom, and all Affinity members.

 

Thank you for looking at the shape images.

I am hoping to engage artists and designers who can use shape images in their workflow; please contact me if you are interested.

 

vonDoom, your comment identifies the presence of pixel-size square shapes in some images; here is an explanation; I hope this is helpful:

 

The method does not combine areas of same colour; in fact, the method works with the greyscale image, determining colours after the shapes have been determined.

 

The main task is to determine polygons (i.e. shapes) having boundaries each of which has constant lightness (i.e. constant gray value). These polygons may be highly irregular in shape; their coordinates are expressed to sub pixel accuracy (1/10th of a pixel for the images in the Gallery). It is these polygons that are the "shapes" in ShapeVision.

 

But as you noticed, some images do have clusters of pixel-sized squares all filled with the same colour; here's why: For some of the constant-boundary polygons (i.e. polygons having boundary of constant lightness), the boundary is not a line, but instead has nonzero area (of constant lightness) -- it is these nonzero-area boundary regions that are filled with pixel-size squares. This fill-by-pixel-size-squares occurs only in some images, as needed, but this is not the dominant mode of operation for ShapeVision.

 

Martin

 

 

Share this post


Link to post
Share on other sites

Hi, Martin,

 

There have been many request made to add in a tracing function for Affinity. Because there is not one, I have used various tracing apps. Yours seems to be as good or better than those I've tried.

 

I played w. the rasberry canes file using AD. The file opened and could be altered w/o incident.

 

I spent some time at the web site trying various options. There were more than I could spend time trying to use. For my creative creative purposes, none were quite realistic enough to translate an image. Some of the sample images when drawn in the coarse setting range did give good results turning the image into a more schematic representation. As you mention, the process was developed for engineering. I would probably use the coarse methods to start simple line drawings that held close to the base image. 

 

Can you offer some images translated for engineering use? From what I've seen on this forum, there are users who make schematic, not creative, images. Perhaps show some transforms of geometric shapes.


iMac 27" Retina, c. 2015: OS X 10.11.5: 3.3 GHz I c-5: 32 Gb,  AMD Radeon R9 M290 2048 Mb

iPad 12.9" Retina, iOS 10, 512 Gb, Apple pencil

Huion WH1409 tablet

Share this post


Link to post
Share on other sites

Thank you gdenby, I would be happy to try to identify an appropriate coarseness that would suit your style of work.

Look for a post from me in a day or two.

If you would like to suggest some sample images, that would be great.

The engineering images that I have are medical, with a concentration in ophthalmology; their image parse may or may not be useful for you.

Or cat and dog images ...

Thanks for your interest!

Martin

 

Share this post


Link to post
Share on other sites

Hello again gdenby and all,

 

As promised, here are some images with coarser settings. Press the green buttons to see each. Please let me know if these satisfy your curiosity, or whether there are other tests that I can do for you that might be useful. For example, please suggest some images for me to try.

 

By the way, I notice that when downloading shape images by right-click, some browsers sometimes attach the wrong file extension. The extension should be .svg, but I have seen both .svgz and .svg.css -- in these cases you can rename the file to .svg either in the browser dialog or after download. I.e. the file is actually an svg file.

 

From the ShapeVision Gallery:
http://d3p13rqcqmn2zd.cloudfront.net/1027091723447502565/index.html
http://d3p13rqcqmn2zd.cloudfront.net/860385316933440383/index.html
http://d3p13rqcqmn2zd.cloudfront.net/19470188640419895/index.html
http://d3p13rqcqmn2zd.cloudfront.net/619611254918745180/index.html

 

Cats & dogs -- images I use to test with deep learning:
http://d3p13rqcqmn2zd.cloudfront.net/80277021170244841/index.html 
http://d3p13rqcqmn2zd.cloudfront.net/563806558909854815/index.html

 

Martin

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...

×
×
  • Create New...

Important Information

Please note the Annual Company Closure section in the Terms of Use. These are the Terms of Use you will be asked to agree to if you join the forum. | Privacy Policy | Guidelines | We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.