Hi,
Thanks for sharing this great project!
I have a question regarding the capabilities of the DART framework. Is it possible to use an encoder–decoder engine within DART to generate segmentation masks directly (e.g., for pixel-wise prediction tasks)?
More specifically:
Does the current implementation support encoder–decoder architectures for mask generation?
If not, would it be feasible to adapt the pipeline to support this use case?
Are there any examples or recommendations for integrating mask prediction into the existing framework?
My use case involves generating spatial masks rather than only bounding boxes or classification outputs, so I want to understand whether this aligns with the current design of DART.
Thanks in advance for your help!
Best regards,
sphuang
Hi,
Thanks for sharing this great project!
I have a question regarding the capabilities of the DART framework. Is it possible to use an encoder–decoder engine within DART to generate segmentation masks directly (e.g., for pixel-wise prediction tasks)?
More specifically:
Does the current implementation support encoder–decoder architectures for mask generation?
If not, would it be feasible to adapt the pipeline to support this use case?
Are there any examples or recommendations for integrating mask prediction into the existing framework?
My use case involves generating spatial masks rather than only bounding boxes or classification outputs, so I want to understand whether this aligns with the current design of DART.
Thanks in advance for your help!
Best regards,
sphuang