facebookresearch / facebookresearch/sonata

Instance Segmentation

Open
#30 2 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
795
Forks
56
PR merge metrics
No merged PRs in 30d

Description

Hi!

I've successfully implemented semantic segmentation training and testing with SONATA on ScanNet 20 but I noticed in the paper that SONATA also achieves strong instance segmentation results. I'm now trying to extend my implementation to instance segmentation but I'm unsure about the approach?

Do you have any recommendations on:
What instance segmentation head architecture works best with SONATA? Also is there any data prep that I need to do in particular?

I'd really appreciate any insights you could share. Great project btw - thank you so much for your work and being so generous with releasing the code/models!

-Alex

Contributor guide

Open the contributing guide

Research direction

The issue names no files or tests. Start by locating the existing SONATA semantic-segmentation training and testing entry points and reviewing the paper's instance-segmentation setup for ScanNet 20; done means a documented, reproducible path for the requested head architecture and data preparation.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.