Dominant Mask Regions

Documentation

• Class name: SaltMaskDominantRegion
• Category: SALT/Masking/Filter
• Output node: False

This node focuses on identifying and isolating the dominant region within a given set of masks based on a specified threshold. It effectively highlights the most prominent area in an image mask, making it useful for tasks that require focus on significant mask regions.

Input types

Required

• masks
  • The input masks on which the dominant region detection is to be performed. These masks are crucial for determining the area of interest within the images.
  • Comfy dtype: MASK
  • Python dtype: torch.Tensor
• threshold
  • A threshold value to distinguish the dominant region within the masks. It plays a pivotal role in defining what constitutes the 'dominant' area by setting a cutoff intensity value.
  • Comfy dtype: INT
  • Python dtype: int

Output types

• MASKS
  • Comfy dtype: MASK
  • The output is a tensor of masks with the dominant region highlighted. This is significant for applications needing to focus on or manipulate the primary area within the masks.
  • Python dtype: torch.Tensor

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class SaltMaskDominantRegion:
    @classmethod
    def INPUT_TYPES(cls):
        return {
                    "required": {
                        "masks": ("MASK",),
                        "threshold": ("INT", {"default":128, "min":0, "max":255, "step":1}),
                    }
                }

    CATEGORY = f"{NAME}/Masking/Filter"

    RETURN_TYPES = ("MASK",)
    RETURN_NAMES = ("MASKS",)

    FUNCTION = "dominant_region"

    def dominant_region(self, masks, threshold=128):
        if not isinstance(threshold, list):
            threshold = [threshold]
        regions = []
        for i, mask in enumerate(masks):
            mask_pil = mask2pil(mask.unsqueeze(0))
            region_mask = MaskFilters.dominant_region(mask_pil, int(threshold[i if i < len(threshold) else -1]))
            region_tensor = pil2mask(region_mask)
            regions.append(region_tensor)
        regions_tensor = torch.cat(regions, dim=0)
        return (regions_tensor,)