Marvin: A Minimalist GPU-only N-Dimensional ConvNets Framework
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
examples example for ImageDataLayer Sep 19, 2016
models Update model JSON files to reflect new MemoryDataLayer Dec 23, 2015
python/marvin Move the python package to a separate folder Nov 21, 2015
tools Fix plotNet: ensure nameofResponse is a cell array of char vectors Mar 21, 2018
.gitignore realtime webcam demo Sep 14, 2016
CHANGELOG Move ActivationLayer initialization to Malloc Apr 8, 2016
CMakeLists.txt compilation and readme upgrade to cudnn 5.1 Sep 16, 2016
LICENSE I'm alive! Nov 9, 2015 compilation and readme upgrade to cudnn 5.1 Sep 16, 2016 compilation and readme upgrade to cudnn 5.1 Sep 16, 2016
marvin.hpp bug fix Jan 28, 2017


Marvin is a GPU-only neural network framework made with simplicity, hackability, speed, memory consumption, and high dimensional data in mind.


Download CUDA 7.5 and cuDNN 5.1. You will need to register with NVIDIA. Below are some additional steps to set up cuDNN 5.1. NOTE We highly recommend that you install different versions of cuDNN to different directories (e.g., /usr/local/cudnn/vXX) because different software packages may require different versions.

LIB_DIR=lib$([[ $(uname) == "Linux" ]] && echo 64)
echo LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDNN_LIB_DIR >> ~/.profile && ~/.profile

tar zxvf cudnn*.tgz
sudo cp cuda/$LIB_DIR/* $CUDNN_LIB_DIR/
sudo cp cuda/include/* /usr/local/cudnn/v5.1/include/




  1. Prepare data: run examples/mnist/prepare_mnist.m in Matlab
  2. Train a model: run ./examples/mnist/ in shell
  3. Visualize filters: run examples/mnist/demo_vis_filter.m in Matlab

Tutorials and Documentation

Please see our website at


The following is the citation of the current version of Marvin. Note that the reference may change in the future when new contributors join the project.

      title        = {Marvin: A minimalist {GPU}-only {N}-dimensional {ConvNet} framework},
      author       = {Jianxiong Xiao and Shuran Song and Daniel Suo and Fisher Yu},
      howpublished = {\url{}},
      note         = {Accessed: 2015-11-10}


Marvin stands on the shoulders of others who have open-sourced their work. You can find the source code of their projects along with license information below. We acknowledge and are grateful to these developers and researchers for their contributions to open source.