The add-on uses machine learning to compare object descriptors.
Sometimes, it can be really difficult to find an object in your Blender library. This Ray Cast Visual Search (RCVP) plug-in is made to help you with that. Just choose something that looks a bit like what you need, press the button, and you have everything that can pass for what you've sculpted.
The system processes 3D models obtained using RGB-D sensors (depth-sensing devices that work in association with an RGB sensor camera,) and then selects and classifies objects and their auto-completion based on visual search.
To implement the visual search algorithm, the researchers got a modification of the ray casting 3D-shape feature extraction algorithm. They also investigated the possibility of using deep learning architectures based on convolution mechanisms to solve the classification problem. According to the authors, auto-completion is made faster and more accurate thanks to the information about the object class obtained during the classification stage.
Simply put, the chosen object's descriptor is compared to the ones in the library. Similar objects have similar descriptors, identical objects have identical descriptors. The size and position don't matter. The result seems impressive, and the plug-in – useful, but it's not yet available to the public.
If you'd like to get updates, join the waitlist here. Also, read about the projects here don't forget to join our new Reddit page, our new Telegram channel, follow us on Instagram and Twitter, where we are sharing breakdowns, the latest news, awesome artworks, and more.