The dataset has been established to enable comparative studies on gait analysis, especially the problems of gait index estimation and abnormal gait detection. The dataset includes 9 gaits that are normal (symmetric) walking gait and 8 simulated abnormal (asymmetric) ones. Since the dataset contains 3 synchronized data types (point cloud, skeleton, frontal silhouette), this is appropriate for assessment on many gait-related methods. This page briefly describes the dataset and provides the download links.
The dataset was acquired by the system containing:
Acquired walking gaits: There were 9 volunteers in our dataset, in which each subject performed 9 walking gaits with different levels of symmetry. Each walking gait was acquired in 1200 consecutive frames.
Notation | Description |
N | normal walking gait |
L|5cm L|10cm L|15cm L|4kg |
padding a sole with a thickness of 5 cm
under left foot padding a sole with a thickness of 10 cm under left foot padding a sole with a thickness of 15 cm under left foot attaching a weight (4 kilograms) to left ankle |
R|5cm R|10cm R|15cm R|4kg |
padding a sole with a thickness of 5 cm
under right foot padding a sole with a thickness of 10 cm under right foot padding a sole with a thickness of 15 cm under right foot attaching a weight (4 kilograms) to right ankle |
Full dataset: 9 subjects × 9 gaits × (1200 clouds + 1200 skeletons + 1200 silhouettes)
Default data separation:
The data included in this dataset can be used, free of charge, for research and educational purposes. The dataset belongs to the Image Processing Laboratory, DIRO, University of Montreal. Copying and/or redistribution without reference as well as any unauthorized commercial use are prohibited. Any researcher reporting results which use this dataset is requested to cite the following paper:
@article{Nguyen20183dreconstruction,
author = {T. N. Nguyen and H. H. Huynh and J. Meunier},
title = {3D Reconstruction With Time-of-Flight Depth Camera and Multiple Mirrors},
journal = {IEEE Access},
year = {2018},
volume = {6},
pages = {38106-38114},
doi = {10.1109/ACCESS.2018.2854262},
ISSN = {2169-3536}
}
Point cloud:
Skeleton: skeletons.zip (45 MB)
Silhouette: silhouettes.zip (66 MB)
Normalized cylindrical histogram: normhists.zip (34 MB) (unzip password: wUz7EcH9xG)
Details of these histograms can be found in our papers: JIVP, PAALast update: August 06, 2018