1from __future__ import annotations
 2from typing import List
 3from dataclasses import dataclass
 4from fuzzywuzzy import fuzz
 5
 6
 7CONF_THRESHOLD = 60
 8# word predictions with lower confidence will be filtered out
 9
10
11@dataclass
12class PredictedWord:
13    __slots__ = 'confidence', 'text'
14    confidence: int
15    text: str
16
17
18class PredictedFrame:
19    index: int  # 0-based index of the frame
20    words: List[PredictedWord]
21    confidence: int  # total confidence of all words
22    text: str
23
24    def __init__(self, index, pred_data: str):
25        self.index = index
26        self.words = []
27
28        block = 0  # keep track of line breaks
29
30        for l in pred_data.splitlines()[1:]:
31            word_data = l.split()
32            if len(word_data) < 12:
33                # no word is predicted
34                continue
35            _, _, block_num, *_, conf, text = word_data
36            block_num, conf = int(block_num), int(conf)
37
38            # handle line breaks
39            if block < block_num:
40                block = block_num
41                self.words.append(PredictedWord(0, '\n'))
42
43            if conf >= CONF_THRESHOLD:
44                self.words.append(PredictedWord(conf, text))
45
46        self.confidence = sum(word.confidence for word in self.words)
47        self.text = ''.join(word.text + ' ' for word in self.words).strip()
48
49    def is_similar_to(self, other: PredictedFrame, threshold=60) -> bool:
50        if len(self.text) == 0 or len(other.text) == 0:
51            return False
52        return fuzz.ratio(self.text, other.text) >= threshold
53
54
55class PredictedSubtitle:
56    frames: List[PredictedFrame]
57
58    def __init__(self, frames: List[PredictedFrame]):
59        self.frames = [f for f in frames if f.confidence > 0]
60
61    @property
62    def text(self) -> str:
63        if self.frames:
64            conf_max = max(f.confidence for f in self.frames)
65            return next(f.text for f in self.frames if f.confidence == conf_max)
66        return ''
67
68    @property
69    def index_start(self) -> int:
70        if self.frames:
71            return self.frames[0].index
72        return 0
73
74    @property
75    def index_end(self) -> int:
76        if self.frames:
77            return self.frames[-1].index
78        return 0