SeC Video Segmentation
SeCVideoSegmentation
Concept-driven video object segmentation using Large Vision-Language Models for visual concept extraction. Provide visual prompts (points/bbox/mask) and SeC automatically understands the object concept for robust tracking.
Pack: Comfyui-SecNodes
custom_nodes.Comfyui-SecNodes
Inputs (13)
| Name | Type | Required |
|---|---|---|
| model | SEC_MODEL | required |
| frames | IMAGE | required |
| positive_points | STRING | optional |
| negative_points | STRING | optional |
| bbox | BBOX | optional |
| input_mask | MASK | optional |
| tracking_direction | COMBO | optional |
| annotation_frame_idx | INT | optional |
| object_id | INT | optional |
| max_frames_to_track | INT | optional |
| mllm_memory_size | INT | optional |
| offload_video_to_cpu | BOOLEAN | optional |
| auto_unload_model | BOOLEAN | optional |