ComfyUI Node
YOLOE-26 Instance Masks
Segment with a text prompt and return one mask per detected instance, with JSON metadata mapping each output mask to its source image and detection.
YOLOE-26 Instance Masks
- model
- image
- instance_masks
- instance_metadata_json
- count
◄promptperson►
◄conf0.10►
◄iou0.70►
◄max_det300►
◄mask_threshold0.50►
◄imgsz640►
CategoryYOLOE26
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| model | YOLOE_MODEL | — | |
| image | IMAGE | — | |
| prompt | STRING | person | Text prompt for open-vocabulary segmentation. Separate multiple classes with commas. |
| confopt | FLOAT | 0.100–1 | Detection confidence threshold. |
| iouopt | FLOAT | 0.700–1 | IoU threshold used by Ultralytics inference. |
| max_detopt | INT | 3001–1000 | Maximum number of detections returned per image. |
| mask_thresholdopt | FLOAT | 0.500–1 | Threshold used when binarizing instance masks. Recent Ultralytics releases already return binary masks, in which case values below 1.0 have no additional effect and exactly 1.0 empties every mask. |
| imgszopt | INT | 64064–2048 | Inference image size. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| instance_masks | MASK | — |
| instance_metadata_json | STRING | — |
| count | INT | — |