Deep Thermal Imaging Dataset

深度热成像数据集由两个主要数据集组成： DeepTherm I（室内材料）- 使用 15 种室内材料创建数据集 DeepTherm I，其中包含 14,860 个处理过的热图像（每个单独类别的平均数据计数：990.7，SD= 425.9；每个变量的每种材料的 400-600 张图像）。该数据集是通过在具有不同照明水平（亮/暗）、有/无空调、不同位置（在地板或桌子上）和不同视角的房间中记录热图像序列创建的（图 5）。从不同角度和不同距离（10 到 50 厘米，从相机镜头到材料）收集空间温度模式。在大约 3 周内收集了五次数据。 DeepTherm II（户外材料） - 17 种户外材料被定位。数据收集过程产生了 DeepTherm II 数据集，其中包括 26,584 个标记的热图像。从每种材料收集的空间热模式的平均数量为 1563.8（SD=295.3；每种材料每种条件下约 300-500 张图像）。

The thermal imaging dataset is composed of two main subsets: DeepTherm I (Indoor Materials) and DeepTherm II (Outdoor Materials).

DeepTherm I (Indoor Materials): This dataset was developed using 15 types of indoor materials, containing a total of 14,860 processed thermal images. The average number of samples per individual category is 990.7, with a standard deviation (SD) of 425.9; each material has 400–600 images per variable. The dataset was created by capturing thermal image sequences in rooms with varying lighting conditions (bright/dim), with or without air conditioning, at different placements (on the floor or table) and from diverse viewing angles (Figure 5). Spatial temperature patterns were collected from various angles and distances ranging from 10 to 50 cm between the camera lens and the material. Data collection was conducted five times over approximately three weeks.

DeepTherm II (Outdoor Materials): A total of 17 types of outdoor materials were utilized for data acquisition, resulting in the DeepTherm II dataset which includes 26,584 labeled thermal images. The average number of collected spatial thermal patterns per material is 1563.8, with an SD of 295.3; each material has approximately 300–500 images per experimental condition.