What about portrait-type image that has studio lighting quality? How would that work? I admit I could have done a bit more tweaking on the colour palette, but I was satisfied with the results as I wanted to see how the algorithm would section out the various features. It processed the image quite well indeed. I think it would be suitable for a poster of a portrait with minor tweaks if you decided to make one of yourself.
Test Subject 5: Christoper Reeves as Superma n. I was having fun at this point just throwing images in to see what would work. Again, my selection of this piece was also with the premise of if I had to really hang it as a decorative piece. I needed something that has a cool factor. The results were not perfect, especially around the insignia, where I wanted the yellow parts to stand out a lot mor.
There would also be some fixing needed with some red parts. The great value about this is that on the first page, it will tell you exactly how many pieces of each element are needed based on the colors you selected. And the final part of the process is to just hit the Bricklink XML button to copy the parts you need to order via Bricklink. Deb: I really liked the idea of Lego Art when it was released, since it reminded me of an assignment from my computer vision class back in college.
I realised that I could use a simplified variation of the algorithm of what was used in the original assignment, followed by a custom post-processing step this is marked as step 4 on the website.
My favourite part of the project as it exists today is still how well the algorithm is able to run due to the pixelated nature of the images—I lucked out in a way that let me make the tool accessible for free to anyone who wants to give it a shot, without having to install any software. The algorithm has limitations, and depending on the input image and the parts available, may not always work very well. Ultimately, the constraints of the medium is what makes Lego art beautiful, and the tool itself is not very useful without a human choosing an image, and helping the algorithm navigate the constraints in a creative way to make something amazing.
TBB: How long did you take to code this? Deb: Where I got lucky was that due to the nature of the problem, the resolution of the images I would have to deal with after the preprocessing step would be very low you can only realistically achieve so much with Lego studs.
This, alongside the way the postprocessing step scales meant that I would not have to worry as much about performance and that I would be able to run this directly in the browser. Since no servers would have to be involved, and I was not dealing with a bunch of other complex infrastructure, the application itself took around 5 hours to build, and a few minutes to deploy.
I put in a couple of hours some days later to address some feature requests, such as more resolutions and custom input creation. Did you see that, and what do you think of it? Tell us a bit about yourself Deb? What do you do? And can you tell us a bit more about your favourite LEGO theme? Deb: I was a pretty big Lego fan growing up, though that kind of fell off in high school and college, mostly due to time constraints.
I got back into it after college, though I need to limit myself because unfortunately I am now cursed with only being able to store so much. I like to combine my hobbies with my knowledge of code and algorithm design, so as I started getting back into Lego, the idea of combining Lego art with computer vision seemed like a pretty natural fit.
It would be so convenient if Lego was able to combine the algorithm from step 3 of the tool with Pick A Brick, or a similar custom service to allow people to create and order their own art pieces directly from Lego, or if Lego just created a Lego art set with a lot of pieces representing different skin tones, paired with an app that uses the full Lego Art Remix algorithm to allow people to easily make pictures of their friends and family from the pieces already in the box.
The beauty of these is that the proof of concept already exists. We wish you the best! It would be interesting to see images that incorporate elements such as metallicity and transparency, and of course, pieces that glow in the dark are cool no matter how old you are.
It can only mean good things if you or someone else now can even expand on it as more ideas develop. All the best in your future projects! We were thrilled to see some of our suggestions make an impact on the development. I did one using photoshop and studio 2. Worked pretty well! Export your final instruction book as high resolution, print-quality images or PDF, or as small-sized images suitable for sharing online.
Web Lic is currently in an early beta stage, and improving quickly. You can follow its progress, ask questions or open bug reports on its GitHub page. Or email me any questions, feedback or bug reports directly. Web Lic is intended to be a complete replacement for the aging and no longer maintained desktop version of Lic. It includes all of the features that were in desktop Lic, more or less, plus some features that would have been impossible to add to desktop Lic, like single steps that span multiple pages, or steps within steps.
This time around, it's architected in a much more simple and maintainable way. And it has unit tests. Web Lic is built using Vue. Models and parts are rendered with Three.
This is a purely client side application; there is no server side component, beyond a web server to serve the initial content, of course. Try It Now. Auto layout Automatically layout the steps of your instruction book to nicely fill each page. Launch Lic works best in Chrome.
0コメント