CASIA-IVA-Lab
prompt process for initial mask
Is possible to create a prompt_process.general_mask(mask), in which mask can be an initial mask and not only points or bounding box(es)?
Currently, the prompt_process only supports points and bounding boxes. However, it would be highly beneficial to extend its functionality to accept an initial mask as well. This would enable users to provide a more flexible and comprehensive mask, allowing for a broader range of use cases.
We have also though about this. Maybe converting the mask to multiple points or using the minimum bounding rectangle of the mask are two of the feasible ways. If you have better strategies, we welcome your furthur detailed suggestions.
Thank you for your quick reply. I need to further study the FastSAM process to provide more insightful suggestions. One potential approach could involve identifying key points within the masked regions of the original image, similar to the SIFT method, and using them as reference points for further analysis.
Thank you for the discussion. Current FastSAM use the location matching method to get the masks from the "everything" masks. Your method seems works better for the SAM-like methods.
When entering a prompt box, fastsam outputs a mask; May I ask if a prompt box can be used as input to output multiple masks within the box area?
