← All blueprints

SAM3 Image/Video Segmentation

Detect and segment objects in images with AI agents

Build an object detection and segmentation system that identifies, classifies, and precisely isolates objects in images. An agent combines object detection models with SAM3 segmentation to deliver both bounding boxes and pixel-accurate masks with class labels.

Stack

EigenForge Agent ForgeObject detection model (YOLO, DETR)SAM3 for segmentationClassification model

Implementation

  1. 1

    Set up object detection

    Deploy an object detection model as an agent tool. Configure confidence thresholds, class filters, and NMS parameters for your use case.

  2. 2

    Chain detection with segmentation

    The agent takes detection bounding boxes and uses them as SAM3 prompts to generate pixel-accurate segmentation masks for each detected object.

  3. 3

    Add classification and attributes

    Extend beyond basic classes. The agent can classify detected objects with fine-grained labels and extract attributes like color, size, and condition.

  4. 4

    Implement spatial analysis

    Build tools for analyzing spatial relationships between detected objects — proximity, overlap, containment, and relative positioning.

  5. 5

    Deploy for your use case

    Configure for your domain: retail (product detection), manufacturing (defect detection), medical (anomaly detection), or security (object tracking).

What You Get

  • Combined object detection and pixel-accurate segmentation
  • Fine-grained classification with object attributes
  • Spatial relationship analysis between detected objects
  • Adaptable to any domain with agent-driven configuration

Ready to build this?

Join the Waitlist