YOLO Segmentation (RMBG) is a ComfyUI node in the 🧪AILab/🧽RMBG category from ComfyUI-RMBG. Inputs: images, model, mask_count, select_mask_index. Outputs…

A sophisticated ComfyUI custom node engineered for advanced image background removal and precise segmentation of objects, faces, clothing, and fashion elements. This tool leverages a diverse array of models, including RMBG-2.0, INSPYRENET, BEN, BEN2, BiRefNet models, SAM, and GroundingDINO, while also incorporating a new feature for real-time background replacement and enhanced edge detection for improved accuracy.

ComfyUI-RMBG

images

model

mask_count

select_mask_index

conf

classes

annotation_mode

background

background_color

mask_blur

mask_offset

invert_output

device

max_det

retina_masks

agnostic_nms

unload_model

AILab_Yolo

IMAGE

ANNOTATED_IMAGE

MASK

MASK_LIST

YOLO Segmentation (RMBG)

Name	Type	Default	Description
images	IMAGE		—
model	COMBO	yolo26m-seg (Medium / 54.8MB)	YOLO model weights stored under /tmp/ComfyUI/models/ultralytics (presets download automatically).
mask_count	COMBO	all	Merge this many detections. 'all' merges everything (or just the selected index when specified).
select_mask_indexopt	COMBO	none	1-based index of the first mask to keep. Use 'none' to start from the first detection.
confopt	FLOAT	0.250–1	Confidence threshold forwarded to Ultralytics.
iouopt	FLOAT	0.450–1	IOU used during NMS.
classesopt	STRING		Comma list or ranges of class IDs; empty keeps every class.
annotation_modeopt	COMBO	Mask & Box	Display mode for ANNOTATED_IMAGE output.
backgroundopt	COMBO	Alpha	Background type for cutout IMAGE output.
background_coloropt	COLORCODE	#222222	Background color when background is set to Color.
mask_bluropt	INT	00–64	Apply Gaussian blur to mask edges.
mask_offsetopt	INT	0-64–64	Expand/Shrink mask boundary (positive = expand, negative = shrink).
invert_outputopt	BOOLEAN	false	Invert the mask output.
deviceopt	COMBO	auto	Force a device or auto-detect CUDA → MPS → CPU.
max_detopt	INT	3001–1000	Maximum detections per image.
retina_masksopt	BOOLEAN	true	Use high-resolution masks (Ultralytics retina_masks flag).
agnostic_nmsopt	BOOLEAN	false	Enable class-agnostic NMS.
unload_modelopt	BOOLEAN	false	Unload model from VRAM after execution.

YOLO Segmentation (RMBG)

Inputs (18)

Outputs (4)