search

MIT-SPARK / Kimera-Semantics

Real-Time 3D Semantic Reconstruction from 2D data
BSD 2-Clause "Simplified" License
645 stars 135 forks source link

Choosing a neural network for semantic segmentation #51

Open amburkoff opened 3 years ago

amburkoff commented 3 years ago

Can you please tell me which neural networks of segmentation do you recommend for use with kimera? What are the requirements for it? How to choose the right one? I collected a real dataset from my robot and launched it on kimera, now I need to select a neural network for semantic segmentation and run semantics together with kimera.

ToniRV commented 3 years ago

Hi @amburkoff , Mask-RCNN is the easiest/fastest to use, but any that outputs semantic labels on a per-pixel basis will do. We are planning to release a how-to wiki page to make this process easier.

nbansal90 commented 3 years ago

Hi @amburkoff , Mask-RCNN is the easiest/fastest to use, but any that outputs semantic labels on a per-pixel basis will do. We are planning to release a how-to wiki page to make this process easier.

Hey @ToniRV , Do we have the wiki ready for integrating off the shelf segmentation model (Example Mask-RCNN) with the Kimera . Please share if Yes; Otherwise a small and approx (step wise) write-up about the required set up would be much appreciated. Thank you