grexai / supercut

Readme for Unsupervised image registration inference

This repository is for multimodal microscopy image registration

Install instructions

Setup a python enviroment

Create a virutal enviroment with python version 3.6

Activate the virtualenv, and run

pip install  -r requirements.txt

Download models

Download models for cut and SuperPoint, and place them into "./models/cut" and ".models/sp/" respectively

Images

Use the example images in the Images/A and Images/B folder

or download the full datasets from https://zenodo.org/record/8162985 and preprocess them

Run

Run

Start the pipeline with arguments: Cut model path: where latest_net_G.pb is located SuperPoint model path, where saved_model.pb is located Image A path: in the article referred as modality 1 Image B path: in the article reffered as modality 2

Example

python3 run_pipeline.py "./models/cut/cut_unaligned_resize/" "./models/sp/sp_v6/" "./Images/A/p1_wA1_t1_m9_c1_z0_l1_o0_1.png" "Images/B/p1_wA1_t1_m9_c1_z0_l1_o0_1.png"