open-mmlab / mmsegmentation

OpenMMLab Semantic Segmentation Toolbox and Benchmark.
https://mmsegmentation.readthedocs.io/en/main/
Apache License 2.0
8.32k stars 2.63k forks source link

Support for Pascal Context-459 and ADE20K-847 #3367

Open develop-productivity opened 1 year ago

develop-productivity commented 1 year ago

Describe the feature

Motivation Recently, a series of open-vocabulary semantic segmentation works have emerged. For example: Scaling Open-Vocabulary Image Segmentation with Image-Level Labels, Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models, Open-vocabulary semantic segmentation with mask-adapted clip, etc. These work typically in Pascal VOC (VOC), Pascal Context-59 (PC-59), Pascal Context-459 (PC-459), ADE20K-847 (ADE-847), ADE20K-150 (ADE-150).

I found that this project does not have code to support PC-459, ADE-847 version data sets. Can we consider adding these two versions in future updates?

Ex1. It is convenient for developers to happily study open-vocabulary semantic segmentation issues based on mmseg. Ex2. There is a recent paper:

Related resources third party code for prepare the datasets can be found ADE-847 PC-459. I find detection2 supports these two versions.

xiexinch commented 1 year ago

Hi @develop-productivity, Thanks for your interest, we will be adding to mmseg soon!