When representing a neural field, instead of having to learn and store one large field for the entire spatial area of interest, hybrid approaches exist which combine features organized in a data structure (e.g. grid) with a small neural net to produce the final result. #CVPR2022 
There are lots of choices for the data structure that organizes the feature vectors: grids, point clouds, meshes…
And even more choices!
Key takeaway: There’s not one single best approach. Each method has its own pros/cons for quality, speed, ease-of-use, etc. The decision comes down to your target application, and picking the best tool for the job.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
