Unverified Commit 6f8fb0b3 authored by dmMaze's avatar dmMaze Committed by GitHub
Browse files

Merge pull request #19 from dmMaze/textlayout 

Automated textlayout
parents d424cf7d cb68b231

ballontranslator/data/pkusegscores.json

0 → 100644
+1 −0
Original line number Diff line number Diff line 
{"n": {"n": 0.0622, "t": 0, "s": 0, "f": 0.0405, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.1921, "a": 0, "z": 0, "d": 0.0891, "p": 0.0258, "c": 0.0228, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "t": {"n": 0.0736, "t": 1, "s": 0, "f": 0.027, "m": 0, "q": 0, "b": 0, "r": 0.172, "v": 0.2027, "a": 0, "z": 0, "d": 0.1426, "p": 0.0384, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "s": {"n": 0.0578, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.3442, "a": 0, "z": 0, "d": 0.102, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "f": {"n": 0.0501, "t": 0, "s": 0, "f": 0, "m": 0.0305, "q": 0, "b": 0, "r": 0.0302, "v": 0.2821, "a": 0, "z": 0, "d": 0.0829, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "m": {"n": 0.2381, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0.4929, "b": 0, "r": 0, "v": 0.0675, "a": 0.0355, "z": 0, "d": 0.0297, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "q": {"n": 0.4815, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0.0329, "v": 0.0943, "a": 0.0826, "z": 0, "d": 0.0447, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "b": {"n": 0.3708, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.026, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0.0207, "vx": 0, "ad": 0, "an": 0}, "r": {"n": 0.1064, "t": 0, "s": 0, "f": 0, "m": 0.0341, "q": 0, "b": 0, "r": 0.0441, "v": 0.2954, "a": 0.027, "z": 0, "d": 0.1562, "p": 0.0429, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "v": {"n": 0.1081, "t": 0, "s": 0, "f": 0, "m": 0.0432, "q": 0, "b": 0, "r": 0.1591, "v": 0.1757, "a": 0.0224, "z": 0, "d": 0.036, "p": 0.0421, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "a": {"n": 0.1464, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0333, "a": 0.0281, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "z": {"n": 0.0737, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.1117, "a": 0.0306, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "d": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0.0313, "v": 0.5905, "a": 0.1047, "z": 0, "d": 0.1317, "p": 0.0626, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "p": {"n": 0.2263, "t": 0.0201, "s": 0.034, "f": 0, "m": 0.0314, "q": 0, "b": 0, "r": 0.4427, "v": 0.0835, "a": 0.0232, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0.0318, "ns": 0.0314, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "c": {"n": 0.1367, "t": 0, "s": 0, "f": 0, "m": 0.0253, "q": 0, "b": 0, "r": 0.2692, "v": 0.2089, "a": 0.0321, "z": 0, "d": 0.108, "p": 0.0514, "c": 0.0296, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0.0274, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "u": {"n": 0.5251, "t": 0, "s": 0, "f": 0, "m": 0.043, "q": 0, "b": 0, "r": 0.0581, "v": 0.0781, "a": 0.0424, "z": 0, "d": 0.0276, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0.0414, "vx": 0, "ad": 0, "an": 0}, "y": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "e": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "o": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0.0412, "q": 0, "b": 0, "r": 0, "v": 0.1031, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "i": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0205, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "l": {"n": 0.0289, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0433, "a": 0, "z": 0, "d": 0.0409, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "j": {"n": 0.2866, "t": 0, "s": 0, "f": 0.0258, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0789, "a": 0, "z": 0, "d": 0.0304, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0.1251, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0.1348, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "h": {"n": 0.5789, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0.0351, "v": 0.0526, "a": 0.1228, "z": 0, "d": 0.0526, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0.0351, "vx": 0, "ad": 0, "an": 0}, "k": {"n": 0.0236, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0.0236, "v": 0.2505, "a": 0, "z": 0, "d": 0.1234, "p": 0.0526, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "g": {}, "x": {}, "w": {"n": 0.0859, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0.2504, "v": 0.1699, "a": 0, "z": 0, "d": 0.1199, "p": 0.0597, "c": 0.1215, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0.0376, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "nr": {"n": 0.1627, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.1824, "a": 0, "z": 0, "d": 0.0654, "p": 0.0393, "c": 0.0417, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0.1234, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "ns": {"n": 0.2905, "t": 0, "s": 0.0305, "f": 0.0397, "m": 0.0257, "q": 0, "b": 0, "r": 0, "v": 0.1452, "a": 0, "z": 0, "d": 0.0317, "p": 0, "c": 0.0274, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0.03, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "nt": {"n": 0.2051, "t": 0, "s": 0, "f": 0.0897, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.2436, "a": 0, "z": 0, "d": 0, "p": 0.0385, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0.0256, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0.0513, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "nx": {"n": 1.0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0, "a": 0, "z": 0, "d": 0, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "nz": {"n": 0.3944, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0646, "a": 0, "z": 0, "d": 0.0347, "p": 0, "c": 0.0221, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "vd": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.7857, "a": 0, "z": 0, "d": 0.119, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "vn": {"n": 0.2322, "t": 0, "s": 0, "f": 0.0323, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.0619, "a": 0, "z": 0, "d": 0.0467, "p": 0, "c": 0.033, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0.0201, "vx": 0, "ad": 0, "an": 0}, "vx": {"n": 0.0396, "t": 0, "s": 0, "f": 0, "m": 0.0223, "q": 0, "b": 0, "r": 0.1609, "v": 0, "a": 0.052, "z": 0, "d": 0.0223, "p": 0, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0.4728, "vx": 0, "ad": 0, "an": 0.047}, "ad": {"n": 0, "t": 0, "s": 0, "f": 0, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.844, "a": 0.0336, "z": 0, "d": 0.0519, "p": 0.0528, "c": 0, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}, "an": {"n": 0, "t": 0, "s": 0, "f": 0.0246, "m": 0, "q": 0, "b": 0, "r": 0, "v": 0.1258, "a": 0, "z": 0, "d": 0.0912, "p": 0, "c": 0.0629, "u": 1, "y": 1, "e": 0, "o": 0, "i": 0, "l": 0, "j": 0, "h": 0, "k": 0, "g": 0, "x": 0, "w": 1, "nr": 0, "ns": 0, "nt": 0, "nx": 0, "nz": 0, "vd": 0, "vn": 0, "vx": 0, "ad": 0, "an": 0}}
 
 No newline at end of file

ballontranslator/data/testpacks/manga/original3.jpg

0 → 100644
+1.48 MiB
Loading image diff...

ballontranslator/dl/ocr/model_32px.py

+5 −1
Original line number Diff line number Diff line
@@ -571,7 +571,11 @@ class OCR32pxModel: 
        for blk_idx, textblk in enumerate(textblk_lst):

            for ii in range(len(textblk)):

                textblk_lst_indices.append(blk_idx)

                regions.append(textblk.get_transformed_region(img, ii, self.text_height))

                region = textblk.get_transformed_region(img, ii, self.text_height)

                h, w = region.shape[:2]

                if w > 3064:    # positional embedding requires width <= 3072

                    region = region[:, :3064]

                regions.append(region)

                region_idx += 1

        # regions = [textblk.get_transformed_region(img, idx, self.text_height) for idx in range(len(textblk))]

        perm = range(len(regions))

ballontranslator/dl/textdetector/textblock.py

+55 −27
Original line number Diff line number Diff line
@@ -3,9 +3,10 @@ import numpy as np 
from shapely.geometry import Polygon

import math

import copy

from utils.imgproc_utils import union_area, xywh2xyxypoly, rotate_polygons

import cv2



from utils.imgproc_utils import union_area, xywh2xyxypoly, rotate_polygons, color_difference



LANG_LIST = ['eng', 'ja', 'unknown']

LANGCLS2IDX = {'eng': 0, 'ja': 1, 'unknown': 2}
@@ -34,14 +35,14 @@ class TextBlock(object): 
                       bold: bool = False,

                       underline: bool = False,

                       italic: bool = False,

                       alignment: int = -1,

                       _alignment: int = -1,

                       alpha: float = 255,

                       rich_text: str = "",

                       _bounding_rect: List = None,

                       accumulate_color = True,

                       default_stroke_width = 0.2,

                       font_weight = 50, 

                       target_lang: str = "",

                       _target_lang: str = "",

                       **kwargs) -> None:

        self.xyxy = [int(num) for num in xyxy]                    # boundingbox of textblock

        self.lines = [] if lines is None else lines     # polygons of textlines
@@ -77,9 +78,8 @@ class TextBlock(object): 
        self.alpha = alpha

        self.rich_text = rich_text

        self.line_spacing = line_spacing

        # self.alignment = alignment

        self._alignment = alignment

        self._target_lang = target_lang

        self._alignment = _alignment

        self._target_lang = _target_lang



        self._bounding_rect = _bounding_rect

        self.default_stroke_width = default_stroke_width
@@ -116,16 +116,20 @@ class TextBlock(object): 
        norm_h = np.linalg.norm(middle_pnts[:, 1] - middle_pnts[:, 3])

        return norm_v / norm_h



    def center(self):

    def center(self) -> np.ndarray:

        xyxy = np.array(self.xyxy)

        return (xyxy[:2] + xyxy[2:]) / 2



    def min_rect(self, rotate_back=True):

    def unrotated_polygons(self) -> np.ndarray:

        angled = self.angle != 0

        center = self.center()

        polygons = self.lines_array().reshape(-1, 8)

        if angled:

            polygons = rotate_polygons(center, polygons, self.angle)

        return angled, center, polygons

    

    def min_rect(self, rotate_back=True) -> List[int]:

        angled, center, polygons = self.unrotated_polygons()

        min_x = polygons[:, ::2].min()

        min_y = polygons[:, 1::2].min()

        max_x = polygons[:, ::2].max()
@@ -135,8 +139,17 @@ class TextBlock(object): 
            min_bbox = rotate_polygons(center, min_bbox, -self.angle)

        return min_bbox.reshape(-1, 4, 2)



    def normalizd_width_list(self) -> List[float]:

        angled, center, polygons = self.unrotated_polygons()

        width_list = []

        for polygon in polygons:

            width_list.append((polygon[[2, 4]] - polygon[[0, 6]]).sum())

        width_list = np.array(width_list)

        width_list = width_list / np.sum(width_list)

        return width_list.tolist()



    # equivalent to qt's boundingRect, ignore angle

    def bounding_rect(self):

    def bounding_rect(self) -> List[int]:

        if self._bounding_rect is None:

        # if True:

            min_bbox = self.min_rect(rotate_back=False)[0]
@@ -166,13 +179,6 @@ class TextBlock(object): 
        direction = 'v' if self.vertical else 'h'

        src_pts = np.array(self.lines[idx], dtype=np.float64)



        if self.language == 'eng' or (self.language == 'unknown' and not self.vertical):

            e_size = self.font_size / 3

            src_pts[..., 0] += np.array([-e_size, e_size, e_size, -e_size])

            src_pts[..., 1] += np.array([-e_size, -e_size, e_size, e_size])

            src_pts[..., 0] = np.clip(src_pts[..., 0], 0, im_w)

            src_pts[..., 1] = np.clip(src_pts[..., 1], 0, im_h)



        middle_pnt = (src_pts[[1, 2, 3, 0]] + src_pts) / 2

        vec_v = middle_pnt[2] - middle_pnt[0]   # vertical vectors of textlines

        vec_h = middle_pnt[1] - middle_pnt[3]   # horizontal vectors of textlines
@@ -198,7 +204,14 @@ class TextBlock(object): 
    def get_text(self):

        if isinstance(self.text, str):

            return self.text

        return ' '.join(self.text).strip()

        text = ''

        for t in self.text:

            if text and t:

                if text[-1].isalpha() and t[0].isalpha():

                    text += ' '

            text += t



        return text.strip()



    def set_font_colors(self, frgb, srgb, accumulate=True):

        self.accumulate_color = accumulate
@@ -239,11 +252,8 @@ class TextBlock(object): 
            return 0

        lines = self.lines_array()

        if len(lines) == 1:

            return 0

        angled = self.angle != 0

        polygons = lines.reshape(-1, 8)

        if angled:

            polygons = rotate_polygons((0, 0), polygons, self.angle)

            return 1

        angled, center, polygons = self.unrotated_polygons()

        polygons = polygons.reshape(-1, 4, 2)

        

        left_std = np.std(polygons[:, 0, 0])
@@ -259,10 +269,8 @@ class TextBlock(object): 


    @property

    def stroke_width(self):

        var = np.array([self.fg_r, self.fg_g, self.fg_b]) \

            - np.array([self.bg_r, self.bg_g, self.bg_b])

        var = np.abs(var).sum()

        if var > 40:

        diff = color_difference(*self.get_font_colors())

        if diff > 15:

            return self.default_stroke_width

        return 0
@@ -331,6 +339,8 @@ def examine_textblk(blk: TextBlock, im_w: int, im_h: int, eval_orientation: bool 
    blk.angle = rotation_angle

    if vertical:

        blk.angle -= 90

    if abs(blk.angle) < 3:

        blk.angle = 0

    blk.font_size = font_size

    if eval_orientation:

        blk.vertical = vertical
@@ -479,6 +489,24 @@ def group_output(blks, lines, im_w, im_h, mask=None, sort_blklist=True) -> List[ 
    final_blk_list += merge_textlines(scattered_lines['ver'])

    if sort_blklist:

        final_blk_list = sort_textblk_list(final_blk_list, im_w, im_h)



    for blk in final_blk_list:

        if blk.language == 'eng':

            num_lines = len(blk.lines)

            if num_lines == 0:

                continue

            # blk.line_spacing = blk.bounding_rect()[3] / num_lines / blk.font_size

            resize_ratio = 1.1

            expand_size = max(int(blk.font_size * 0.1), 2)

            rad = np.deg2rad(blk.angle)

            shifted_vec = np.array([[[-1, -1],[1, -1],[1, 1],[-1, 1]]])

            shifted_vec = shifted_vec * np.array([[[np.sin(rad), np.cos(rad)]]]) * expand_size

            lines = blk.lines_array() + shifted_vec

            lines[..., 0] = np.clip(lines[..., 0], 0, im_w-1)

            lines[..., 1] = np.clip(lines[..., 1], 0, im_h-1)

            blk.lines = lines.astype(np.int64).tolist()

            blk.font_size += expand_size

            

    return final_blk_list



def visualize_textblocks(canvas, blk_list:  List[TextBlock]):

ballontranslator/ui/canvas.py

+4 −0
Original line number Diff line number Diff line
@@ -78,6 +78,7 @@ class Canvas(QGraphicsScene): 
    finish_erasing = Signal(StrokeItem)

    delete_textblks = Signal()

    format_textblks = Signal()

    layout_textblks = Signal()



    begin_scale_tool = Signal(QPointF)

    scale_tool = Signal(QPointF)
@@ -401,11 +402,14 @@ class Canvas(QGraphicsScene): 
            menu = QMenu()

            delete_act = menu.addAction(self.tr("Delete"))

            format_act = menu.addAction(self.tr("Apply font formatting"))

            layout_act = menu.addAction(self.tr("Auto layout"))

            rst = menu.exec_(event.screenPos())

            if rst == delete_act:

                self.delete_textblks.emit()

            elif rst == format_act:

                self.format_textblks.emit()

            elif rst == layout_act:

                self.layout_textblks.emit()

    

    def on_hide_canvas(self):

        self.alt_pressed = False