Commit 405db031 authored by dmMaze's avatar dmMaze
Browse files

update ocr test

parent 80b5a66c

dl/ocr/mit48px_ctc.py

+2 −0
Original line number Diff line number Diff line 
# modified from https://github.com/zyddnys/manga-image-translator/blob/main/ocr/model_48px_ctc.py



import torch

import torch.nn as nn

import torch.nn.functional as F

dl/textdetector/textblock.py

+0 −5
Original line number Diff line number Diff line
@@ -159,11 +159,6 @@ class TextBlock(object): 


    def to_dict(self, extra_info=False):

        blk_dict = copy.deepcopy(vars(self))

        # if not extra_info:

            # blk_dict.pop('distance')

            # blk_dict.pop('weight')

            # blk_dict.pop('vec')

            # blk_dict.pop('norm')

        return blk_dict



    def get_transformed_region(self, img, idx, textheight) -> np.ndarray :

tests/test_ocr.py

+0 −16
Original line number Diff line number Diff line
@@ -34,25 +34,9 @@ def test_mit48px(proj: ProjImgTrans, device: str = 'cpu', chunk_size: int = 16): 




if __name__ == '__main__':

    # setup_params = OCRMIT32px.setup_params

    # setup_params['device']['delect'] = 'cpu'

    # ocr = OCRMIT32px(**setup_params)

    # img_path = r'data/testpacks/textline/ballontranslator.png'

    # img = imread(img_path)

    # rst = ocr.run_ocr(img)



    # proj_dir = r'data/testpacks/manga'

    # proj = ProjImgTrans(proj_dir)

    # for imgname in proj.pages:

    #     img_path = osp.join(proj.directory, imgname)

    #     img = imread(img_path)

    #     blk_list = proj.pages[imgname]

    #     ocr.ocr_blk_list(img, blk_list)

    # proj.save()

    manga_dir = 'data/testpacks/manga'

    comic_dir = 'data/testpacks/testpacks/eng'

    comic_dir2 = 'data/testpacks/testpacks/eng2'

    manga_proj = ProjImgTrans(manga_dir)

    # comic_proj = ProjImgTrans(comic_dir)



    test_mit48px(manga_proj, 'cpu', 16)

ui/stylewidgets.py

+3 −5
Original line number Diff line number Diff line 
from PyQt5.QtWidgets import QGraphicsOpacityEffect, QFrame, QWidget, QComboBox, QLabel, QSizePolicy, QDialog, QProgressBar, QMessageBox, QMenu, QSpacerItem, QPushButton, QHBoxLayout, QVBoxLayout, QToolButton, QSplitter, QStylePainter, QStyleOption, QStyle, QSlider, QProxyStyle, QStyle, QStyleOptionSlider, QColorDialog

from PyQt5.QtWidgets import QGraphicsDropShadowEffect, QGraphicsOpacityEffect, QFrame, QWidget, QComboBox, QLabel, QSizePolicy, QDialog, QProgressBar, QMessageBox, QVBoxLayout, QStylePainter, QStyleOption, QStyle, QSlider, QProxyStyle, QStyle, QStyleOptionSlider, QColorDialog

from PyQt5.QtCore import Qt, QPropertyAnimation, QEasingCurve, QPointF, QRect, pyqtSignal, QSizeF, QObject, QEvent

from PyQt5.QtGui import QFontMetrics, QMouseEvent, QShowEvent, QWheelEvent, QResizeEvent, QKeySequence, QPainter, QTextFrame, QTransform, QTextBlock, QAbstractTextDocumentLayout, QTextLayout, QFont, QFontMetrics, QColor, QTextFormat, QTextCursor, QTextCharFormat, QTextDocument

from typing import List, Union, Tuple
@@ -92,20 +92,18 @@ class ProgressMessageBox(QDialog): 
        super().__init__(*args, **kwargs)

        self.setWindowFlags(Qt.WindowType.FramelessWindowHint)

        self.setSizePolicy(QSizePolicy.Expanding, QSizePolicy.Expanding)



        self.setModal(True)

        layout = QVBoxLayout(self)

        self.label = QLabel(self)

        

        self.detect_bar = TaskProgressBar('detect', self.tr('Detecting: '), self)

        self.ocr_bar = TaskProgressBar('ocr', self.tr('OCR: '), self)

        self.inpaint_bar = TaskProgressBar('inpaint', self.tr('Inpainting: '), self)

        self.translate_bar = TaskProgressBar('translate', self.tr('Translating: '), self)



        layout = QVBoxLayout(self)

        layout.addWidget(self.detect_bar)

        layout.addWidget(self.ocr_bar)

        layout.addWidget(self.inpaint_bar)

        layout.addWidget(self.translate_bar)



        layout.setSpacing(0)

        layout.setContentsMargins(20, 10, 20, 30)

utils/io_utils.py

+0 −2
Original line number Diff line number Diff line
@@ -35,8 +35,6 @@ def find_all_imgs(img_dir, abs_path=False): 
    return imglist



def imread(imgpath, read_type=cv2.IMREAD_COLOR):

    # img = cv2.imread(imgpath, read_type)

    # if img is None:

    if not osp.exists(imgpath):

        return None

    img = cv2.imdecode(np.fromfile(imgpath, dtype=np.uint8), read_type)