Chapter 7: Feature detection and matching

• The HPatches dataset and benchmark (Balntas, Lenc et al. 2020) is often used to evaluate new feature detectors and descriptors.

• The Image Matching Benchmark (Jin, Mishkin et al. 2021) is also widely used and has associated workshops.

• Visual localization datasets such as Aachen Day-Night (Sattler, Maddern et al. 2018) are also often used.

• Pointers to datasets for evaluating instance retrieval algorithms can be found in Zheng, Yang, and Tian (2018).

• Non-semantic image segmentation (splitting an image into “reasonable pieces” without labeling their content) is not widely studied any more. Pointers to classic datasets such as the Berkeley Segmentation Dataset and Benchmark (Martin, Fowlkes et al. 2001) can be found in the first edition of this book (Szeliski 2010, Appendix C.1).