README.md

datasets

Folder containing scripts to build, download and store datasets.

1. Prepare nvidia-docker container

Follow container README instructions.

2. Open project in nvidia-docker container

Run container if it is not running:

cd $HOME/ViT-PCM
bash container/main.sh 

3. Download and convert datasets

- PascalVOC 2012 download, build and convert in tfrecords

Once you ran tensorflow nvidia-docker container as in step 2, inside container type:

cd datasets
source download_and_convert_voc12.sh

- MS-COCO 2014 (also 2017) download, build and convert in tfrecords

Once you ran tensorflow nvidia-docker container as in step 2, inside container type:

cd datasets
source download_and_convert_coco.sh -y 2014 # support also 2017
# use flag '--to-voc12 true' to use only PascalVOC12 category set