acerta-abide

Deep learning using the ABIDE data

Environment Setup

In order to run the deep learning model, you need to install docker and nvidia-docker.

Data preparation

The first step is to download the dataset:

nvidia-docker run -it --rm \
    -v $(realpath data):/opt/acerta-abide/data \
    acerta-abide \
    python download_abide.py

This command will download the preprocessed CC-200 dataset from Amazon S3.

And compile the dataset into CV folds and by experiment.

nvidia-docker run --rm \
    -v $(realpath data):/opt/acerta-abide/data \
    acerta-abide \
    python prepare_data.py \
        --whole \
        --male \
        --threshold \
        --folds 10 \
        cc200 \
    2> /dev/null

Model training

nvidia-docker run --rm \
    -v $(realpath data):/opt/acerta-abide/data \
    acerta-abide \
    python nn.py \
        --whole \
        --male \
        --threshold \
        --folds 10 \
        cc200 \
    2> /dev/null

Model evaluation

nvidia-docker run --rm \
    -v $(realpath data):/opt/acerta-abide/data \
    acerta-abide \
    python nn_evaluate.py \
        --whole \
        --male \
        --threshold \
        --folds 10 \
        cc200 \
    2> /dev/null