Dataset loading scripts for the 3D IQ test task (3D-IQTT) dataset.
Example:
This is the companion code repository for the 3D-IQTT dataset. You need to download the dataset first. No worries, you don't need to sign anything.
The dataset consists of 3 files:
- training, 46GB compressed, 79GB uncompressed*
- test, 7.9GB
- validation, 7.9GB
The command for decompressing the training dataset is
unxz --verbose 3diqtt-v2-train.h5.xz
The --verbose is important to get a progress bar, because this takes ~1-3h.
git clone https://github.com/fgolemo/3D-IQTT.git
cd 3D-IQTT
pip install -e . # you might have to use "sudo"
Check out the file scripts/1-load-train-data.py. That file contains a self-contained example on how to load the different datasets. You just have to change the directoy in line 10.
On macosx, before using the script, make sure you have a ~/.matplotlib/matplotlibrc file with:
backend: TkAgg