Skip to content

Latest commit

 

History

History
24 lines (15 loc) · 991 Bytes

README.md

File metadata and controls

24 lines (15 loc) · 991 Bytes
Raw

Surface match dataset

This dataset contain data which shows value of crossing between two images.
More detail information about this purposes you can read in this article.

progress

How to use

  1. Clone this repo.
  2. Unzip /data/data.zip to get json files with data.
  3. Run /scripts/prepare_data.py from own folder to get npz file for train your neural network.

Structure of this repo

/data

This folder contains images and json files which describe data about this images.
You need unzip /data/data.zip to get json files. It require in script /scripts/prepare_data.py.

/scripts/prepare_data.py

This script creates npz file (/train-data/data_224x224.npz) which you can use to train neural network and images (/train-data/images/*.png) if you want read it from disk directly.
Npz file will contain 2.1 GB and folder with images around 1.5 GB.