FLIR Systems has announced the availability of its first European thermal imaging regional dataset and the third in a series of thermal imaging datasets for machine vision testing. Designed to help researchers, developers and auto manufacturers enhance and accelerate work on safety, advanced driver assistance-systems (ADAS), automatic emergency braking (AEB) and autonomous vehicle (AV) systems, the dataset features thousands of annotated thermal images of daytime, nighttime and various weather scenarios from six cities across Europe.
The new European regional dataset includes image annotations captured in London (England), Paris (France), Madrid, Toledo, Granada and Malaga (Spain). Collected with FLIR thermal and RGB cameras to identify various attributes of city landscapes, the dataset contains a total of 3,895 annotated thermal images to increase testing and evolving convolutional neural networks (CNN). The images were taken in a variety of driving environments in each city, including various lighting and weather conditions. Similar to the previously released thermal imaging datasets for San Francisco, California and Santa Barbara, California, the European dataset includes image annotations for cars, other vehicles types, people, bicycles and signs.
The two previous datasets have been downloaded 5,000 times, and the new images and videos from the European dataset will be added to the growing library of available video and images to further speed up CNN training of thermal data. By incorporating the thermal dataset into automotive design and development, researchers and developers can develop more reliable systems due to the ability of thermal imaging to classify objectives.