Model zoo

With net_download command and the model zoo server, NiftyNet provides convenient access to the shared trained/untrained networks. Trained networks can be used directly (as part of a workflow or for performance comparisons), fine-tuned for different data distributions (e.g. a different hospital’s images), or used to initialize networks for other applications (i.e. transfer learning). Untrained networks or conceptual blocks can be used within new networks.

The following sections introduce:

  • net_download used to download model and data,
  • the user’s “niftynet home” folder as the output directory of net_download (also as the default folder of NiftyNet).

net_download

The command is available for both pip-installed and source code repository users (the source code repository users should replace net_download with python net_download.py):

usage: net_download [-h] [-r] [-v] sample_id [sample_id ...]

Download NiftyNet sample data

positional arguments:
  sample_id      Identifier string(s) for the example(s) to download

optional arguments:
  -h, --help     show this help message and exit
  -r, --retry    Force data to be downloaded again
  -v, --version  show program's version number and exit

For a concrete example,

net_download highres3dnet_brain_parcellation_model_zoo

will automatically take the following two steps:

[config]
version = 1.0

[data]
local_id = OASIS
url = http://cmic.cs.ucl.ac.uk/platform/niftynetexamples/OASIS.tar.gz?dl=1
action = expand
destination = data

[weights]
local_id = highres3dnet_brain_parcellation
url = https://www.dropbox.com/s/nxg2ixs9rh1p9ri/highres3dnet_brain_parcellation_weights.tar.gz?dl=1
action = expand
destination = models

[network_inference_config]
local_id = highres3dnet_brain_parcellation
url = https://www.dropbox.com/s/r2q08q1kkd534p4/highres3dnet_brain_parcellation_config.tar.gz?dl=1
action = expand
destination = extensions
  • parse each section of the .ini file, download [data], [weights], and [network_inference_config] respectively from the specified url, to user’s $NIFTYNET_HOME/[destination]/[local_id] where $NIFTYNET_HOME is specified in the following section. These destination directories are designed for different types of data, possible values of destination are data, models, extensions. Specifically,
    • data directory stores example image inputs.
    • models directory stores trained model weights
    • extensions directory stores Python implementation of networks, loss functions, new applications, etc.

Depending on their availability, some model zoo entries do not contain data, or trained weight, or both.

Global-settings

The global NiftyNet configuration is read from ~/.niftynet/config.ini. NiftyNet attempts to load this file for the global configuration.

  • If it does not exist, NiftyNet will create a default one.
  • If it exists but cannot be read (e.g., due to incorrect formatting):
    • NiftyNet will back it up with a unique string;
    • create a default one.

Currently the minimal version of this file will look like the following:

[global]
home = ~/niftynet

The home key specifies the root folder (referred to as $NIFTYNET_HOME from this point onwards) to be used by NiftyNet for storing and locating user data such as downloaded models, and new networks implemented by the user. This setting is configurable, and upon successfully loading this file NiftyNet will attempt to create the specified folder, if it does not already exist.

On first run, NiftyNet will also attempt to create the NiftyNet extension module hierarchy (extensions.*), that allows for the discovery of user-defined networks. This hierarchy consists of the following:

  • $NIFTYNET_HOME/extensions/ (folder)
  • $NIFTYNET_HOME/extensions/__init__.py (file)
  • $NIFTYNET_HOME/extensions/network/ (folder)
  • $NIFTYNET_HOME/extensions/network/__init__.py (file)

To completely uninstall NiftyNet, please manually remove ~/.niftynet and $NIFTYNET_HOME folders.