THE KTH-TIPS2 database

1 Background This document provides a brief Users' Guide to the KTH-TIPS2 image database (KTH is the abbreviation of our university, and TIPS stands for Textures under varying Illumination, Pose and Scale). The KTH-TIPS2 provides a considerable extension to our previous database of images of materials — KTH-TIPS. The guide describes which materials are contained in the database (Section 2), how images were acquired (Section 3) and subsequently cropped to remove the background (Section 4), and we also discuss some non-ideal artifacts, like poor focus, in some pictures (Section 5). The objective with this database was to provide a more satisfactory means of evaluating algorithms for classifying materials. As we argued in [4, 1], a very relevant task is to recognise categories of materials such as " wood " or " wool " as opposed to one particular physical sample. The KTH-TIPS2 contains four physical samples of 11 different materials. In addition, it is frequently necessary to perform recognition in unstructured environments. Thus the database provides images with variations in scale as well as variations in pose and illumination, following on from the philosophy of the KTH-TIPS, and in part the CUReT image database [2]. The 11 materials in the KTH-TIPS2 database are all present also in the CUReT database [2], which opens the possibility of conducting experiments on a combination of the two databases. The cropped database is freely available on the internet [5]. Those interested in the full-size images should contact Eric Hayman ([email protected]). The database was first presented and used in [1]. The KTH-TIPS2 database contains images of 11 materials (Table 1 and Figure 1), each of which are also present in the CUReT database [2], and six of which were also included in the first KTH-TIPS database [3]. Each of the samples is planar.