ComfyUI nodes for video object segmentation using [a/SAMURAI](https://github.com/yangchris11/samurai) model.

SAMURAI Nodes for ComfyUI

# SAMURAI Refine Node

This node performs video object segmentation using the SAMURAI model.

## Inputs:
- **image**: Input video frames sequence (required)
- **model_name**: SAMURAI model to use (required)
- **resolution**: Maximum resolution for processing (default: 1024)
- **iou_threshold**: Intersection over Union threshold (default: 0.1)
- **box**: Bounding box from SAMURAI Box Input (optional)
- **points**: Point prompts for segmentation (optional)
- **labels**: Labels for point prompts (optional)
- **start_frame**: Starting frame number (optional)

## Outputs:
- **MASK**: Generated segmentation masks
- **frame_number**: Current frame number for sequence

## Models Available:
- sam2.1_hiera_base.pt
- sam2.1_hiera_base_plus.pt
- sam2.1_hiera_large.pt
- sam2.1_hiera_small.pt

## Usage:
1. Connect video and box/points inputs
2. Select desired model and parameters
3. Node will generate masks for the selected object through the video sequence

ComfyUI Node: SAMURAI Refine

Category

Inputs

Outputs

Extension: SAMURAI Nodes for ComfyUI

Run ComfyUI workflows in the Cloud!