The LYNRED Mobility Dataset helps you develop your algorithms
Artificial intelligence (AI) is playing a crucial role in driving the digital transformation of our society, especially in the field of computer vision. This technology enables machines to understand and interpret the visual world in a similar manner to humans, which paves the way for innovative applications in a broad array of sectors, including healthcare, security, and the automotive industry.
However, to be truly effective, these systems require extensive and varied training datasets, which are essential for building robust AI models offering improved accuracy. Aware of these challenges, Lynred is actively committed to powering research in this area by providing thermal image datasets, with the goal of advancing computer vision applications and pushing the boundaries of what technology can achieve.
This dataset comprises three parts covering complementary fields:
LYNRED Mobility Dataset: Multimodal Detection
The LYNRED Multimodal Detection Dataset is specifically designed to support the development of Advanced Driver-Assistance Systems (ADAS) and self-driving vehicles by providing a comprehensive collection of thermal and visible-light RGB data. It comprises 8,000 synchronized and aligned infrared and visible-light RGB images captured in a wide range of environmental conditions, including every season, as well as daytime and nighttime scenes. Researchers and engineers can harness this dataset to develop and test their algorithms in real-world driving scenarios.
| Images | Metadata (JSON) | Camera specifications |
|---|---|---|
| 8,000 visible-light RGB and thermal images with a suggested train test split | 80k labels divided into 9 classes (person, car, bicycle, motorcycle, truck, bus, animal, train, construction machine) | Thermal camera : VGA 640x480 16bits & 8bits |
| Occlusion and temperature metadata | Visible-light RGB camera : SXGA 1280x960 8bits |
LYNRED Mobility Dataset: Stereovision
The LYNRED Stereovision dataset includes everything needed to develop algorithms for image registration, visible-thermal fusion and depth estimation. It contains 43,200 images from six video sequences using two thermal and two visible cameras synchronized by an external trigger in various urban, rural, daytime and night-time scenarios.
| Images | Metadata | Camera specifications |
|---|---|---|
| 6 sequences of synchronized quad-camera images (2 visible-light RGB & 2 thermal cameras) | Camera registration (extrinsic calibration) | Thermal camera : 2 x VGA 640x480 16 bits & 8 bits |
| Visible-light RGB camera : 2 x SXGA 1280x960 8 bits |
LYNRED Mobility Dataset: Range Estimation
The LYNRED Range Estimation Dataset provides a large number of sequences of pedestrians crossing the road at multiple distances, captured from a fixed camera viewpoint inspired by the New Car Assessment Program (NCAP) and National Highway Traffic Safety Administration (NHTSA) scenarios, two renowned car safety performance assessment programs. It can be used to evaluate the detection range of a thermal-based PAEB system up to 250 meters, as well as support a varied range of applications relating to pedestrian detection, self-driving vehicles, and ADAS, which goes beyond the scope of the NCAP or NHTSA test protocols.
| Images | Metadata (CSV) | Camera specifications |
|---|---|---|
| 250+ sequences of pedestrians crossing the road captured simultaneously with a QVGA sensor and a VGA sensor | Pedestrian metadata : height, distance, speed | Thermal cameras : VGA 640x480 16bits & 8bits QVGA 320x240 16bits & 8bits |
| Scenario conditions : rural, urban, day, night, winter, summer | ||
| Environmental conditions : temperature, luminosity |
Our Partners
We would like to thank our partners for their participation and contribution in developing this dataset
Related Products
VGA 17µm
VGA 12µm
QVGA 12µm
Related Publications
Characterization of the infrared sensor capability to identify pedestrians when coupled with a detection algorithm and simulation-based investigation of the potential safety benefits of thermal camera adoption
This paper explores the benefits of visible-thermal fusion, proposes a windshield integration for such a system, and evaluates the minimum performance requirements for a thermal camera compliant with the NHSTA standards, based on a field study of pedestrian detection range
We designed a new approach to visible and thermal fusion for pedestrian detection. Our method uses multi-task learning to train an end-to-end model to perform detection on the multimodal information and on each modality simultaneously
Effect of different infrared processing pipelines on the performance of a pedestrian detection in an urban environment, similar to autonomous driving scenarios
We propose a complete pipeline to process and fuse infrared and visible-light RGB images specialized on driving application
How to cite
LYNRED, LYNRED Mobility Dataset V1 (2025), https://www.lynred.com/lynred-mobility-dataset
