Problematic :
How can multimodal imaging be used to improve the detection of road users?
With recent advances in technology, particularly in advanced driver-assistance systems (ADAS), it has become essential to quickly detect road users in order to reduce the risk of accidents. However, conventional cameras reach their limits when visibility is poor (due to fog, glare, or at night). The combined use of RGB (color) and infrared images improves visibility and detection in all conditions, which represents a major advantage in the development of these systems.
LYNRED-MDS Dataset
It is in this context that the Deep Red Chair, supported by the Grenoble INP Foundation through sponsorship from LYNRED, created the LYNRED-MDS dataset, which combines infrared and RGB imagery to provide more diverse and realistic data for the development of more advanced computer vision systems. Unlike existing datasets, LYNRED-MDS contains 4,000 image pairs, acquired in a wide variety of environments—urban, rural, and mountainous areas—as well as under diverse seasonal and temperature conditions, to better represent the diversity of possible situations on the road.
Image acquisition
The images in the LYNRED-MDS dataset were acquired using two cameras mounted on the roof of a test vehicle. These cameras were positioned as shown in the diagram below and synchronized to capture images at the same moment, at a rate of 30 frames per second. However, only one image is retained every three seconds to maximize scene diversity.
For the object detection task, only the pair of cameras on the left is used. To ensure that images are properly aligned within the region of interest, the sensors are oriented so that objects located at a great distance (in this case, more than 10 meters away) appear aligned across all sensors.
(a) Diagram of the acquisition system.
(b) Photograph of the actual device installed on the vehicle’s roof.
Access to data
The dataset is available for free at this link.
Reference
https://ieeexplore.ieee.org/document/11533396
L. Arbez, J. Matias, X. Brenière, J. Chanussot and R. Phlypo, « Descriptor: LYNRED Mobility Dataset Multimodal Detection Subset (LYNRED-MDS), » in IEEE Data Descriptions, vol. 3, pp. 413-421, 2026, doi: 10.1109/IEEEDATA.2026.3695478
