How to get mask from segmentation model?

[jibby2803]

I've followed the code in #28 and get the results like that

How can I get mask from there?

[ahatamiz]

Hi @jibby2803 Thanks for your interest and inquiry. The above example that you showed only extracts features from the train classification backbone at different stages. If you would like to obtain segmentation mask (output), we kindly request that you look into the semantic segmentation part.

We provide various trained models as listed in the results table.

[jibby2803]

Thank @ahatamiz for your reply

I'm a newbie in Mamba and try to understand and apply mambaVision for a segmentation task for a research project. I have some question to be clarified

1. Is that True: MambaVision segmentation modei is include MambaVision backbone and UperUnet and the checkpoint [here]{https://huggingface.co/nvidia/MambaVision-T-1K} is just params of backbone and use for feature extraction?
   If it True, does this mean I can use the MambaVision backbone as image encoder and replace the UperUnet but any other layer/decoder which can produce mask for image generation task?

2. I've find paper about MambaVision for computer vision tasks and it is mostly applied for medical tasks and not many papers related to downstream task. Why it is?

3. I want to apply MambaVision to segmentation in general domain, are there anything need to be considers?

I will appreciate if you can share me some of your insight about MambaVision after doing research on it!
Thanks for your time!