Iris recognition

Implementation of ZKML iris recognition, inspired by the worldcoin iris recognition pipeline.

How to use

1. Input Generation

Hardcode the path of your iris images in src/main.py (default uses samples)

Then to generate the iris code input ("input.json") for the ONNX model run (it will also perform the matching and ouput the distance in a non proved way) :

python src/main.py

File proving/matching/input.json has been generated.

cd proving/matching

The directory contains :

• network.onnx : the hamming distance onnx model
• input.json : the input of the model
• gen_calibration.py (optionnal, can be executed to generate calibration.json)

2. Proving

With EZKL CLI

Install EZKL then

1. Setup and prove :

ezkl gen-settings 
ezkl compile-circuit
ezkl gen-witness
ezkl gen-srs
ezkl setup 
ezkl prove

With Archon CLI

Install archon then :

1. Create artifact :

archon create-artifact -a matching --data-path input.json --model-path network.onnx

2. Setup and prove :

archon job -a matching gen-settings 
archon job -a matching compile-circuit
archon job -a matching gen-witness
archon job -a matching gen-srs
archon job -a matching setup 
archon job -a matching prove