/cnngait

CNN Gait library for gait-based biometrics

Primary LanguageMATLABGNU General Public License v3.0GPL-3.0

CNN Gait Library for Gait-based Biometrics

Francisco M. Castro and Manuel J. Marin-Jimenez

This library contains support Matlab code for [1] and [2]. If you find useful this code, please, cite [1] or [2].

Prerequisites

  1. MatConvNet library: http://www.vlfeat.org/matconvnet/
  2. Tukey's loss function (for regression): https://github.com/bazilas/matconvnet-deepReg

This code has been tested on Ubuntu 14.04 with Matlab 2013b and Ubuntu 16.04.2 with Matlab 2016b.

Pretrained models

Download from here: ICIP'2017

Sample test data

Some test sequences of the test subject partition from TUM-GAID, normal scenario: download

Quick start

Let's assume that you have placed CNNGait library in folder <cgdir>. Start Matlab and type the following commands:

cd <cgdir>
startup_cnngait
cg_demo_test

References

[1] Deep Multi-Task Learning for Gait-based Biometrics
MJ Marin-Jimenez, F Castro, N Guil, F de la Torre, R Medina-Carnicer
International Conference on Image Processing (ICIP), 2017

[2] Automatic learning of gait signatures for people identification
FM Castro, MJ Marín-Jiménez, N Guil, NP de la Blanca
International Work-Conference on Artificial Neural Networks (IWANN), 257-270