A powered wheelchair detects and follows a caregiver walking beside it. Caregiver recognition is performed by detecting the caregiver feet above the floor. The dataset contains a set of images and annotated labels for a feet recognition application. The camera is placed in the right armrest of the powered wheelchair, tilted down 45 degrees. The camera measures the caregiver's feet walking beside the powered wheelchair. The dataset also includes the training, validation and test files definition for different camera image outputs (Depth, RGB, RGB-D), scenarios and light conditions.

Images are captured by an Intel Realsense D455 depth camera. The dataset includes separate image files for each camera output (Depth, RGB, RGBD) and annotated labels for each frame. Each camera measurement contains 30 seconds of images, recorded at 6 frames/second (180 frames). The total number of images is 6000 (3000 including feet labels and 3000 without). The labelling process has been performed using the software labelImg. Sample scenarios and related frame numbers are defined in the readme.txt file.

Data contains personal data

Language

English

Method and outcome

Sampling procedure

The camera is placed in the right armrest of the powered wheelchair, tilted down 45 degrees. The camera measures the caregiver's feet walking beside the powered wheelchair. The dataset also includes the training, validation and test files definition for different camera image outputs (Depth, RGB, RGB-D), scenarios and light conditions.

Data format / data structure

Text

Still image

Data collection

Geographic coverage

Administrative information

Responsible department/unit

Department Electronic design

Topic and keywords

Research area

Electrical engineering, electronic engineering, information engineering (Standard för svensk indelning av forskningsämnen 2011)

Robotics (Standard för svensk indelning av forskningsämnen 2011)

Keywords

Deep learning, Yolov4 detector, Object recognition, Depth cameras, Autonomous driving vehicles

Publications

Giménez, Cristian V., Silvia Krug, Faisal Z. Qureshi, and Mattias O’Nils. 2021. "Evaluation of 2D-/3D-Feet-Detection Methods for Semi-Autonomous Powered Wheelchair Navigation" Journal of Imaging 7, no. 12: 255. https://doi.org/10.3390/jimaging7120255
DOI: https://doi.org/10.3390/jimaging7120255

If you have published anything based on these data, please notify us with a reference to your publication(s). If you are responsible for the catalogue entry, you can update the metadata/data description in DORIS.