Running complete experimentsΒΆ

We provide an experiment command that runs training, followed by prediction and evaluation. After running, you will be able to find results from model fitting, prediction and evaluation under a single output directory.

For example, to train a model on a pre-configured datamodule evaluate its performance, outputting predictions and performance curves, run the one of following:

# example classification task using the "pasa" network model
# on the "montgomery" datamodule
$ mednet experiment -vv pasa montgomery
# check results in the "results" folder

# example semantic sementation task using the "lwnet" network model
# on the "drive" datamodule
$ mednet experiment -vv lwnet drive
# check results in the "results" folder

# example object detection task using the "faster-rcnn" network model
# on the "montgomery" (for object detection) datamodule
$ mednet experiment -vv montgomery-detect faster-rcnn
# check results in the "results" folder

You may run the system on a GPU by using the --device=cuda, or --device=mps option.