LATCHESS21: dataset of damaged chessboard lattice points (chessboard features) used to train LAPS detector (grayscale/21x21px)

Large dataset of damaged __chessboard lattice points__ (chessboard features), generated\r\nautomatically by modifying perfect 2D standard 8x8 chessboard grid, then\r\nevaluated on real photographs during detection process.\r\n\r\n__10%__ - perfect cases\r\n__10%__ - warped (random direction)\r\n__50%__ - evaluated during detection (real)\r\n__20%__ - mostly mistaken (objects)\r\n__10%__ - very hard (destroyed)\r\n\r\nIt contains the following subsets:\r\n\r\n- __ok__: correct chessboard lattice points\r\n- __no__: similar structures but not correct (convergent)\r\n\r\n_This dataset was used to train LAPS detector described in our article._