Find every copy of an image, and the evidence of who made it.
Bakau is image fingerprinting and provenance infrastructure: it finds every copy of an image, records who made what, and makes rights legible to software and buyers.
Bakau is early and still finding its form. One person builds and runs it: Joseph Sherrier, a student at Georgia Tech. About
Where to start
For software partners
A matching engine that runs on the device, offline. Licensing and
integration.
For archives and catalogs
An audit of what your catalogue holds, run on your own systems,
with a free diagnostic.
For creators and associations
Where this is heading: provenance at the point of creation and
collective licensing. Creators never pay.
Built today
- Copy matching that survives crops. On 200 public-domain photographs with 19 edits each, it found 99% of copies, with 0 false matches in 3,200,000 pairs of unrelated images. It looks for copies, not look-alikes, and can occasionally join two photographs of one scene by one photographer. The benchmark
- A catalogue audit that runs where the images are. Groups duplicates and variants, reads the creator and rights fields in each file, flags gaps and conflicts, and writes a report, a CSV and JSON. A sample report
- A timestamped record. Every record in a report is a leaf of a Merkle tree whose root is filed with OpenTimestamps. The sample report's filing
- No image is stored. Images are read in memory, turned into digital fingerprints and let go. No image is stored, here or in an audit.
Public preview
Choose up to 10 images, for example a photograph and copies of it that were cropped, resized or re-saved. You will see which are the same picture, what their metadata names, and what an audit would flag. They are read in memory for this one request and never stored.
The preview looks for copies, not look-alikes, and it reports what files say, never who owns them. Questions: outreach@bakau.org.