For open source projects, I participated twice for Google Summer of Codes mentored by OpenCV on deep learning and 3D pose estimation tasks. Additionally, I do demonstrate some research works in terms of 3D animations, for example, there is a video got released on my personal website (https://wangyida.github.io/) for 3D scene understanding from 2D depth image. I’m also interested to build fancy websites for GSoD projects based on similar framework as what I am using for my current 3D vision research (https://wangyida.github.io/).
As for my thesis, I am currently working on 3D indoor scene understanding topics and dealing with 3D CV projects including 3D printing and 3D registration projects which are collaborated with other research groups. My topics are totally matched to OpenSCAD's interest, so it should be a good chance for me to participate for GSoD this year. The initial plan could be focused two topics which are also rather related to each other.
I. Package OpenSCAD for Education & Research Purposes
As users are interested in the way to deploy the generated 3D CAD models in their own areas, it will be good to further show users how OpenSCAD could be used easily and efficiently on their own interests. First of all, for the benefit for the both 3D modeling and 3D printing community, OpenSCAD can be deployed with optional hyper-parameters.
Inspired by the potential topics posted on OpenSCAD’s GSoD sites, I plan to firstly include the following areas: Visualization of Geometry, Math Art, Recursive Functions, Creating 3D shapes and Simple animations for Physics.
Additionally, I can deal with some good 3D visualisation for point clouds, volumetric data and CAD models. I have ever rendered 3D models with good shadows and fine vertexes and use those 3D models for my research papers.
II. Improve the OpenSCAD User Manual
As the OpenSCAD user manual is maintained at Wikibooks which has the benefit that users can easily contribute changes but it does mean less control regarding what's documented and changed. The manual is not well structured and also skips over some important topics.
I plan to explain the possibilities the user interface gives (as opposed to the modelling language which is covered by the Language Manual).
I plan to host our webpage for user manual on local servers or Github pages using Hugo (https://gohugo.io/) framework which is easy to get updated and flexible to get rendered both for desktop and mobile phones. Hugo is one of the most popular open-source static site generators with its amazing speed and flexibility.
With large amount of possible themes (https://themes.gohugo.io/), we can host the whole documents of OpenSCAD with label-based searching approaches or just with tags.
Re: Plans to work with OpenSCAD for GSoD on CAD model related topics
On 21.06.19 23:02, Yida wrote:
> Improve the OpenSCAD User Manual
> And Package it for Education & Research Purposes
Interesting additional topic. So far I can't remember much
feedback of OpenSCAD being used in Research. The closest
is probably the old YouTube videos about reverse engineering
solid models (e.g. STL files) into a tree of primitives
combined by CSG. So basically translating STL to SCAD which
is really not an easy topic.
I've never seen anything more than just those videos which
are about 5 years old by now:
> I can deal with some good 3D visualization for point clouds,
> volumetric data, and CAD models. I have ever rendered 3D
> models with good shadows and fine vertexes and use those 3D> models for my research papers.
That are certainly areas where OpenSCAD is not that well
equipped so far. So it would be interesting to hear what
specific visualization and scenarios you have in mind for
> So I plan to host it using Hugo (https://gohugo.io/) which
> is easy to get updated and flexible to get rendered both for
> desktop and mobile phones. Hugo is one of the most popular
> open-source static site generators with its amazing speed
> and flexibility.
I have not heard of hugo before, but it certainly looks
promising by being open source and actively maintained.
Also I like the use of simple markdown files as main
format, similar to what we use in a simple way for our
GSoC site this year (based on github pages).
It would be nice to see a little bit more details regarding
both the proposed timeline for the time GSoD runs and also
more specifics about the topics that will be covered by the
documentation. With focus on the User Manual I would assume
there's quite a number of ways to approach the task.
There is probably the need to select a couple of features to
cover in depth as with the newer stuff like the Customizer
a full coverage might be too much for the GSoD time.
Re: Plans to work with OpenSCAD for GSoD on CAD model related topics
> Interesting additional topic. So far I can't remember much
> feedback of OpenSCAD being used in Research. The closest
> is probably the old YouTube videos about reverse engineering
> solid models (e.g. STL files) into a tree of primitives
> combined by CSG. So basically translating STL to SCAD which
> is really not an easy topic.
> I've never seen anything more than just those videos which
> are about 5 years old by now:
Yes, similar to this work on cloud-wise visualizations. For example, in
research topics for 3D data, volumetric and point clouds data are used for
data-driven approaches rather than mesh data presented in CAD models. Works
like 3D object completion (https://github.com/Yang7879/3D-RecGAN-extended)
and 3D scene semantic reconstruction (https://github.com/shurans/sscnet)
convert CAD model into volumetric data but still convert them back for a
Basically, their data are resampled as point clouds or volumetric data from
rendered from CAD model, and then for demonstration, they will convert those
data format back into 3D CAD model for demonstration with Microsoft PPT or
Apple Keynote with good shadowing effects and vertex visualization. Those
data processing and demonstration procedures are quite simple but seldom got
mentioned in any publications and open source products because they are not
related to some specific algorithms. I personally have a quite simple tool
targeted on this CAD model to volumetric data to CAD model.
--type 0: ply, 1: 3mf, >=2: both (default: 0)
--pI (REQUIRED) The path of input data. (default: )
--pO The path of output file. (default: output)
--pL The path of label file. (default: colorMap.txt)
--sN The number of input data. (default: 1)
--sW The width of input data. (default: 80)
--sH The height of input data. (default: 48)
--sD The depth of input data. (default: 80)
--s The scale of output file. Default is unit volume. (default: 1)
--sb The space between each cube. (default: 0.8)
--t The number of thread to run. (0: auto detect, >0: the number you
want.) (default: 0)
--bff Flip the surface direction (default: 0)
--bgcm Enalbe the generating mode for color labeling. (default: 0)
--bbb In enable, add 8 vetices on the each cornor of the volume. (default:
--bt In enable, using texture mapping instead of color. (default: 0)
> I have not heard of hugo before, but it certainly looks
> promising by being open source and actively maintained.
> Also I like the use of simple markdown files as main
> format, similar to what we use in a simple way for our
> GSoC site this year (based on github pages).
Yes, Hugo will render the website by converting the markdown file into
specific HTML files with predefined styles. So there are several advantages
by doing this:
1. Developers and document makers do not concentrate on tedious HTML
documents, markdown files are simple enough for inserting images and
2. The rendered websites are stylish yet friendly for all devices like
desktop devices and mobile phones.
3. All styles for subpages and hyperlinks are kept identical with each other
which make it professional and easy to read.
4. open sourced tool.
> It would be nice to see a little bit more details regarding
> both the proposed timeline for the time GSoD runs and also
> more specifics about the topics that will be covered by the
> documentation. With focus on the User Manual I would assume
> there's quite a number of ways to approach the task.
> There is probably the need to select a couple of features to
> cover in depth as with the newer stuff like the Customizer
> a full coverage might be too much for the GSoD time.
I think 3 months are fine to carry this out, I can also maintain it further
after this season if you like it. And there will also be a simple document
describing how to insert materials for all of our developers, so it could be
easy to maintain the website.
Actually, it's also good to follow what we are using now for all documents
and make them with rich information and good explanations:)
Ph.D. student in Technical University of Munich for 3D computer vision and machine learning
Sent from: http://forum.openscad.org/