Thermal Dataset for People with Mobility Restrictions
- Citation Author(s):
-
Xiao Ni
(University of Muenster)
- Submitted by:
- Xiao Ni
- Last updated:
- DOI:
- 10.21227/71zq-r907
- Links:
17 views
- Categories:
- Keywords:
Abstract
We present a thermal imaging dataset specifically designed to support the detection of people with mobility restrictions in urban traffic environments. The dataset was collected at multiple intersections and includes diverse spatial regions such as sidewalks, waiting zones, and pedestrian crossings. Data acquisition was conducted across various times of day (sunrise, morning, afternoon, sunset, night, and dawn) and seasons (spring, summer, autumn, and winter) to capture a broad range of environmental and thermal conditions. A total of 11,196 thermal images are included, with a balanced seasonal distribution (3,086 spring, 2,584 summer, 2,534 autumn, 2,992 winter) to minimize bias in model training and evaluation. This dataset supports research on thermal-based pedestrian detection, especially for people with mobility restrictions.
Instructions:
YOLO Format
Directory Structure
/train/
├── images/
└── labels/
/valid/
├── images/
└── labels/
Annotations
Each image has a corresponding .txt file in the labels/ directory.
Each line in the file follows this format:
<class_id> <x_center> <y_center> <width> <height>
Class Definitions
Defined in the data.yaml file
