search

lxtGH / OMG-Seg

OMG-LLaVA and OMG-Seg codebase [CVPR-24 and NeurIPS-24]
Other
1.32k stars 50 forks source link

multi-label for each point #9

Closed suliangxu closed 4 months ago

suliangxu commented 7 months ago

Thank you for your excellent work! Can OMG-seg predict multi-class for each point? For the human parsing task, we need to segment not only the instance of one person but also the partition of him/her (such as the watch, head, upper body, shoes, and so on). Additionally, how can I fine-tune OMG-seg for such a task? Thank you.

image

lxtGH commented 6 months ago

Hi! @suliangxu Seems like this is an instance part-aware segmentation task. You can refer to these works for your project.

https://github.com/lxtGH/Panoptic-PartFormer

https://github.com/soeaver/awesome-human-parsing

suliangxu commented 6 months ago

@lxtGH Thank you for your reply! Another question: how can I test the model for a single image? For example, I can directly specify the path of the image in test.py.

lxtGH commented 5 months ago

@suliangxu Please wait for a well. I will add one folder on how to use these models.

lxtGH commented 4 months ago

Please check the demo folder.