Nodes/ComfyUI-YogurtSa2VA/Yogurt Sa2VA Image Segmentation V2
ComfyUI Node

Yogurt Sa2VA Image Segmentation V2

Segment one image with a reusable Sa2VA model and optional VITMatte detail refinement.

By yogurt7771·Created 3 months ago·Updated 3 months ago· 0
Yogurt Sa2VA Image Segmentation V2
  • sa2va_model
  • vitmatte_model
  • image
  • text_output
  • masks
segmentation_promptPlease provide segmentation masks for all objects.
threshold0.50
process_detailtrue
detail_erode6
detail_dilate6
black_point0.15
white_point0.99
max_megapixels2.0
CategoryYogurtSa2VA

Inputs (11)

NameTypeDefaultDescription
sa2va_modelYOGURT_SA2VA_MODEL
vitmatte_modelYOGURT_VITMATTE_MODEL
imageIMAGEInput image to segment. Should be in RGB format.
segmentation_promptSTRINGPlease provide segmentation masks for all objects.Text prompt describing what objects to segment in the image.
thresholdFLOAT0.500–1Threshold for converting probability masks to binary masks.
process_detailBOOLEANtrueEnable VITMatte post-processing for smooth edges and fine details.
detail_erodeINT61–255Erosion kernel size for trimap generation.
detail_dilateINT61–255Dilation kernel size for trimap generation.
black_pointFLOAT0.150.01–0.98
white_pointFLOAT0.990.02–0.99
max_megapixelsFLOAT2.00.5–10Max resolution for VITMatte processing.

Outputs (2)

NameTypeDescription
text_outputSTRINGText output from the model
masksMASKGenerated segmentation masks