Commit a4808920 authored by dmMaze's avatar dmMaze
Browse files

improve eng->chi textlayot

parent 3c9f8966

ballontranslator/ui/scenetext_manager.py

+15 −12
Original line number Diff line number Diff line
@@ -493,6 +493,7 @@ class SceneTextManager(QObject): 


        max_central_width = np.inf

        if tgt_is_cjk:

            if ballon_area / text_area > 2:

                if blkitem.blk.text:

                    _, _, brw, brh = blkitem.blk.bounding_rect()

                    br_area = brw * brh
@@ -505,6 +506,9 @@ class SceneTextManager(QObject): 
                        max_central_width = max(normalized_width_list)

                else:

                    resize_ratio = 1.1

            else:

                if ballon_area / text_area < 1.8:   # default eng->cjk font_size = 1.1 * detected_size, because detected eng bboxes are a bit small

                    resize_ratio = 0.9



        if resize_ratio != 1:

            new_font_size = blk_font.pointSizeF() * resize_ratio
@@ -515,7 +519,7 @@ class SceneTextManager(QObject): 
            delimiter_len = int(delimiter_len * resize_ratio)



        if max_central_width != np.inf:

            max_central_width = int(max_central_width * text_w)

            max_central_width = max(int(max_central_width * text_w), 0.8 * region_rect[2])



        padding = pt2px(blk_font.pointSize()) + 20   # dummpy padding variable

        if fmt.alignment == 1:
@@ -535,7 +539,6 @@ class SceneTextManager(QObject): 
                centroid[0] = int(abs_centroid[0] - mask_xyxy[0])

                centroid[1] = int(abs_centroid[1] - mask_xyxy[1])





        new_text, xywh = layout_text(mask, mask_xyxy, centroid, words, wl_list, delimiter, delimiter_len, blkitem.blk.angle, line_height, fmt.alignment, fmt.vertical, 0, padding, max_central_width)

        

        # font size post adjustment

ballontranslator/utils/text_layout.py

+2 −2
Original line number Diff line number Diff line
@@ -148,7 +148,7 @@ def layout_lines_aligncenter( 
            new_len = line.length + wl + delimiter_len

            new_x = centroid_x - new_len // 2

            right_x = new_x + new_len

            if new_x <= 0 or right_x >= bw:

            if new_x <= 0 or right_x >= bw or new_len > max_central_width:

                line_valid = False

            elif mask[pos_y: line_bottom, new_x].sum() > 0 or\

                mask[pos_y: line_bottom, right_x].sum() > 0:
@@ -181,7 +181,7 @@ def layout_lines_aligncenter( 
            new_len = line.length + wl + delimiter_len

            new_x = centroid_x - new_len // 2

            right_x = new_x + new_len

            if new_x <= 0 or right_x >= bw:

            if new_x <= 0 or right_x >= bw or new_len > max_central_width:

                line_valid = False

            elif mask[pos_y: line_bottom, new_x].sum() > 0 or\

                mask[pos_y: line_bottom, right_x].sum() > 0:

ballontranslator/utils/text_processing.py

+17 −5
Original line number Diff line number Diff line
@@ -2,8 +2,9 @@ from typing import List, Tuple 
from tqdm import tqdm

import json



WIDE_MAP = {i: i + 0xFEE0 for i in range(0x21, 0x7F)}

WIDE_MAP[0x20] = 0x3000

HALF2FULL = {i: i + 0xFEE0 for i in range(0x21, 0x7F)}

HALF2FULL[0x20] = 0x3000

HALF2FULL[0x2E] = 0x3002

FULL2HALF = dict((i + 0xFEE0, i) for i in range(0x21, 0x7F))

FULL2HALF[0x3000] = 0x20

FULL2HALF[0x3002] = 0x2E
@@ -23,7 +24,7 @@ def full_len(s: str): 
    Convert all ASCII characters to their full-width counterpart.

    https://stackoverflow.com/questions/2422177/python-how-can-i-replace-full-width-characters-with-half-width-characters 

    """

    return s.translate(WIDE_MAP)

    return s.translate(HALF2FULL)



def half_len(s):

    '''
@@ -125,14 +126,14 @@ def _seg_ch_pkg(text: str) -> List[str]: 
                word_next, tag_next = segments[ii + 1]

                len_next = len(word_next)

                next_valid = True

                if tag_next != 'w' and word_next != '.':    # somehow pkgseg take '.' as 'n'

                if tag_next != 'w':

                    score_next = PKUSEGSCORES[tag][tag_next]

            

            if ii > 0:

                word_prev, tag_prev = words[-1], segments[ii - 1][1]

                len_prev = len(word_prev)

                prev_valid = True

                if tag_prev != 'w' and word_prev[-1] != '.':

                if tag_prev != 'w':

                    score_prev = PKUSEGSCORES[tag_prev][tag]



            append_prev, append_next = False, False
@@ -188,6 +189,13 @@ def seg_ch_pkg(text: str): 
        import pkuseg

        CHSEG = pkuseg.pkuseg(postag=True)



    # pkuseg won't work with half-width punctuations

    fullen_text = full_len(text)

    cvt_back = False

    if fullen_text != text:

        cvt_back = True

        text = fullen_text



    global PKUSEGSCORES

    if PKUSEGSCORES is None:

        with open(PKUSEGPATH, 'r', encoding='utf8') as f:
@@ -204,6 +212,10 @@ def seg_ch_pkg(text: str): 
            result_list.extend(words)

    if len(result_list) > 0:

        result_list = result_list[1:]



    if cvt_back:

        # pkuseg w

        result_list = [half_len(word) for word in result_list]

    return result_list



def seg_text(text: str, lang: str) -> Tuple[List, str]: