A new thermal image dataset for facilitating hotspot detection and identification.
The TIHD Dataset consists of 7361 images. 4017 of these images are images that do not contain hotspots, and are unlabeled. On the other hand, 3344 images of this dataset are labeled as either 'normal' meaning that the image has no real hotspots, or 'anomaly' meaning that the image in question contains either one, two or three hotspots. Of these 3344 labeled images, fake hotspots have been added to 1672 images, and these are the images that are labeled as 'normal'.
The subject of the images in this dataset are different shaped aluminium metal sheets of different sizes. The shapes of the metal sheets are rectangle, rhombus, oval and triangle. These metal sheets were captured from an approximate distance of 2 meters with slightly randomly varying angles, using an InfiRay IRS-PT264 Dual Spectrum Thermal Camera. The dimensions of each shape are as follows:
- Rectangle: 33cm x 55cm
- Rhombus: 26cm (side length)
- Oval: 38cm (major axis length), 32cm (minor axis length)
- Triangle: 38cm (base length), 52cm (perpendicular height from the base center)
The dimensions of each image in the dataset is 1280x1024, and each image is a heatmap.
For getting access to the dataset please fill up the short form at https://forms.gle/YAXsR2A5PeQzX1H16
For any questions, contact Shreyas Goyal at [email protected]