Table 2 Dataset distribution: this table represents the distribution of data for SCI and healthy individuals

From: 3D pose estimation for scalable remote gait kinematics assessment

Datasets

SCI

Healthy

Data source

SCAI-Gait

H3.6M13, HumanEva I28

Individuals

225

7, 4

Gender ratio

171 males, 54 females

6 males, 5 females

Mean age

46.06 ± 17.23 years

30 ± 5 years

Mean height

1.75 ± 0.10 m

1.75 ± 0.10 m

Mean weight

74.81 ± 15.92 kg

70.0 ± 10.0 kg

Trials

2-3 (1 POV)

2 (3 POVs), 3 (3 POVs)

Video resolution

1920 × 1080 (50 FPS), 720 × 576 (25 FPS)

1000 × 1000 (50 FPS), 640 × 480 (25 FPS)

Total frames

46,717 (36,925, 9792)

47,248 (38,881, 8367)

  1. These datasets were used to carry out 3D Pose estimation-based benchmarking. The best 3D Pose Estimation Neural Network model was identified and used for gait multi-variate time series data generation. This was then processed further using a K-Means Clustering method to filter only subjects and trials having gait signal. This was then followed by classification and feature extraction using K-Means classifier and Multi-Layered Perceptron (MLP) classifier.
Back to article page