Skip to content

Make a good NFT marker

NFT (Natural Feature Tracking) recognises an image by its feature points: corners and small high-contrast details. The more distinctive points an image has, spread over the whole surface, the more stable the tracking.

Good targets:

  • lots of detail and texture everywhere (photos, illustrations, busy posters)
  • strong local contrast
  • no large flat areas of one colour
  • not repetitive (tiles, regular patterns and text grids confuse the matcher)
  1. Open the NFT Marker Creator and upload a JPG or PNG.
  2. Check the confidence stars: they estimate how well the image will track.
  3. Press Generate: you get three files, .iset, .fset and .fset3.

Images above about 1000 px per side can take a long time in the browser; for those use a threaded command-line version (Node.js, Python or Docker).

Load the three files in FeatureSet Display (how to read it). Feature points should cover the whole image. Areas with no points are areas the tracker can’t see: if the camera only frames those, tracking is lost.

Print at the size you will use it, on matte paper if you can: glossy paper reflects light and hides features.