01 / REFERENCE
+One object example
02 / TARGET
→A new image
OUTPUT
Every match
Change the reference, not the model
Designed for applications where targets change too often for a conventional dataset and retraining cycle.
More than a mask
Planned outputs include masks, boxes, scores, centers, contours, area, count, and later pose.
Productization underway
The core result is promising. Fine-grained distinction between similar objects and speed are the current focus.
Want to test a use case?
Send one reference image, representative target images, target hardware, and the output your software needs. Early samples help define the public demo.
Share a sample ↗