Prepare ADE20K dataset.¶
ADE20K is a scene-centric containing 20 thousands images annotated with 150 object categories. This tutorial help you to download ADE20K and set it up for later experiments.
You need 2.3 GB free disk space to download and extract this dataset. SSD harddrives are recommended for faster speed. The time it takes to prepare the dataset depends on your Internet connection and disk speed. For example, it takes around 25 mins on an AWS EC2 instance with EBS.
Prepare the dataset¶
We will download and unzip the following files:
The easiest way is to run this script:
If you have already downloaded the above files and unzipped them,
you can set the folder name through
--download-dir to avoid
downloading them again. For example
python ade20k.py --download-dir ~/ade_downloads
How to load the dataset¶
Loading images and labels from ADE20K is straight-forward with GluonCV’s dataset utility:
Get the first sample¶
Visualize data and label¶
Total running time of the script: ( 0 minutes 0.000 seconds)