Dataset Statistics

Overview

The InclusiveVidPose dataset contains a diverse collection of video sequences featuring individuals with various types of limb deficiencies. Below are the key statistics of our dataset.

313 Videos
398 Unique Individuals
327,235 Frames Annotated
308,533 Poses Annotated

Demographic Distribution

Demographic Distribution

Figure 1: Overview of Participant Distribution by Deficiency Type, Gender, and Prosthesis. (a) Distribution of limb-deficiency types, with single-site deficiencies and multi-site cases. (b) Gender Distribution. (c) Distribution of individuals with and without prosthetic limbs.

Limb Deficiency Types

Type Count Percentage
Upper Limb Amputation
Lower Limb Amputation
Congenital Limb Difference
Multiple Limb Differences

Activity Distribution

Activity Distribution

Figure 2: Distribution of different activities in the dataset

Dataset Splits

Split Videos Frames
Training
Validation
Testing

Annotation Statistics

  • Average number of keypoints per frame:
  • Total annotated instances:
  • Average annotation time per video:
  • Inter-annotator agreement: