Algorithms and methods by Schomaker & coresearchers

NrName Publ. Comment Code Data
1 Hinge (pixels, edges, convolution) Schomaker & Bulacu (2004) (by convolution) Slow but not so sensitive if character contours are not complete ANSI C Firemaker data set
2 Hinge (coco contour angles) Schomaker & Bulacu (2004) (along connected-component contours) Fast but more sensitive to broken character contours ANSI C Firemaker data set
3 CO3 Fraglet codebooks Schomaker & Bulacu (2004) Usage histogram of allographic elements using connected-component contours ANSI C Firemaker data set
4 CO3 Fraglet codebooks Schomaker & Bulacu (2004) Usage histogram of allographic elements using connected-component pixel images ANSI C Firemaker data set
nn Junction detection Sheng He & Schomaker (2015) Detection of junction points in characters, document scans C++ --
nn DeepOtsu Sheng He & Schomaker (2019) Iterative, deep-learning based preprocessor as a step before general Otsu C++ --
nn Curvature-free features Sheng He & Schomaker (2017) Features for writer identification, not based on angular but bit-pattern information C++ --
nn Multiple-label guided clustering algorithm for historical document dating and localization Sheng He & Schomaker (2016) A constrained elongated Kohonen map in slices over time for style evolution analysis C++ --
nn Fragnet: Writer identification using deep fragment networks Sheng He & Schomaker (2017) Deep learning method for writer identification C++ --
nn BiNet: U-net based binarisation for historical documents Maruf Dhali & Schomaker (2019) Deep learning method for binarizing manuscripts on papyrus or parchment Python IAA/DSS
nn Delta-N hinge: rotation-invariant features Sheng He & Schomaker (2014) Shape feature for writer identification or other rotation-invariant tasks Python --
nn DTGAN Zhenxing Zhang & Schomaker (2021) Single, end-to-end text-to-image GAN Python MSCOCO,...
nn DiverGAN Zhenxing Zhang & Schomaker (2021) Single, end-to-end text-to-image GAN, enforcing diversity Python MSCOCO,...
nn OptGAN Zhenxing Zhang & Schomaker (2021) Single, end-to-end text-to-image GAN, maximizing probability of good output images Python MSCOCO,...

Under construction, June 2022
this list is far from complete