Juliu<p><span class="h-card"><a href="https://sigmoid.social/@esteng" class="u-url mention" rel="nofollow noopener noreferrer" target="_blank">@<span>esteng</span></a></span> some images for context.</p><p>- colored one: sketches to figure out how to do the one-hot-encoding for shapes of diff size.<br>- grid: gerated wireframes. Only couple of shapes, but lots of positions, sizes.<br>- the actual UX app: It has a lot more shapes, etc. Canvas is wider than single wireframe.<br>- graph: results training on 4x4. But larges sizes and more shapes quickly became much harder.</p><p><a href="https://indieweb.social/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a> <a href="https://indieweb.social/tags/wireframes" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>wireframes</span></a> <a href="https://indieweb.social/tags/neuralnets" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>neuralnets</span></a></p>