中文 · Unified manual · Shared tutorials
BrainNet Viewer displays brain surfaces, networks, surface maps and atlas ROIs through MATLAB Desktop, Python Desktop and Docker/Web. This source tree uses version 2.0.0 for all three editions.
Use packages and image tags from the matching GitHub or NITRC release and check their SHA256SUMS. If the 2.0.0 assets are not yet listed, use the source-installation instructions. The earlier 2.0.0-beta.1 downloads remain a separate release; their Python wheel is 2.0.0b1. Changes describes the updates. Python requires 3.11 or newer; the Docker image includes Python 3.12.
- MATLAB: install and activate MATLAB, set Current Folder to this repository
root and run
app = BrainNetViewer;. Do not recursively add another BrainNet Viewer installation. See getting started. - Python Desktop: follow installation, install
the wheel with its desktop extra and run
python -m brainnet_viewer. MATLAB is not required. - Web: follow Docker installation. The service runs in the background and opens at http://127.0.0.1:1031/index.html. Download the Web package and run its Compose pull/start commands; no source build is required once the matching image is available. Image: mingruixia/brainnet-viewer:2.0.0.
Select your inputs in Data Sources and adjust their appearance in Visualization Settings. Click Apply to render, then use the camera tools to explore a view or Export Figure to save it. Save Scene records the inputs and settings for later use; a preset stores display settings for reuse with other data. The shared manual and tutorials explain the file formats, scientific interpretation and differences between editions.
Read display and statistical interpretation when choosing thresholds, size mappings and anatomical overlays.
The installation includes a surface library, atlas-network examples and tutorial
data. AtlasNetworks
contains AAL90 and fourteen GroupFC pairs, including AAL116, Schaefer1000 and
Brainnetome246. AAL90 volume and vertex-label inputs are in
resources/atlases/AAL90/ (source/MATLAB archive) or atlases/AAL90/
(Python/Docker resource library). Inspect uses AAL/Brodmann lookup in an MNI-like space; verify
registration and coordinate correspondence for your own data. The application
interface is English and is not restricted to human-brain data.
The local Web mode has one shared session and is intended for a trusted user on a private local computer. The supplied launchers test GPU rendering and select the GPU when available, otherwise the CPU. GPU devices must be allocated to the container. Compose without GPU allocation runs on the CPU. Save a portable Scene before stopping or updating the container.
For a trusted team on a Linux LAN server, use server mode, with HTTPS, an access password and one application process per browser session. Use a matching 2.0.0 image or a source build; the earlier beta image does not include server mode.
The local server checks browser requests but does not authenticate operating-system users. Another local process can forge request headers, so this configuration is unsuitable for shared-user computers or public hosting without a separately configured deployment. See security guidance.
Python packaging and Docker build files are in src/python. Create an isolated
environment and install that directory with the desktop or web extra.
A local wheel can be built with python -m pip wheel --no-deps ./src/python.
The Docker guide provides source-build commands.
The default container port is 1031.
Xia M, Wang J, He Y (2013). BrainNet Viewer: A Network Visualization Tool for Human Brain Connectomics. PLoS ONE 8: e68910. https://doi.org/10.1371/journal.pone.0068910
Contact Mingrui Xia: mxia@bnu.edu.cn; mingruixia@gmail.com. Project code uses GPL-3.0-or-later. Retain code and bundled-resource notices.