# see huggingface/project # pip install numpy==1.24.4 opencv-python==4.10.0.84 flask==3.1.0 -i https://mirrors.aliyun.com/pypi/simple/ # paddlex 3.0.0b2 requires numpy==1.24.4 # numpy==1.23.5 # .88 see /root/tools/flask_auto_selection.py # 44c08650fa2e54c6be381530a8dd8b5e 这个有问题 is_debug = False dic_cache = {} from flask import Flask, request, jsonify app = Flask(__name__) import numpy as np import cv2 import json import base64 import os, math, re def save_json(filename, dics): with open(filename, 'w', encoding='utf-8') as fp: json.dump(dics, fp, indent=4, ensure_ascii=False) fp.close() def load_json(filename): with open(filename, encoding='utf-8') as fp: js = json.load(fp) fp.close() return js # convert string to json def jsonparse(s): return json.loads(s, strict=False) # convert dict to string def jsonstring(d): return json.dumps(d, ensure_ascii=False) def show_img(image, target_width=400): # 获取原始图片的宽度和高度 original_height, original_width = image.shape[:2] # 计算缩放比例和目标高度 scale = target_width / original_width target_height = int(original_height * scale) # 等比例缩放图片 resized_image = cv2.resize(image, (target_width, target_height), interpolation=cv2.INTER_AREA) cv2.imshow("green", resized_image) cv2.waitKey(0) return resized_image def get_selections(dir, m5): dir_pplabel = 'pplabel' if not os.path.exists(dir_pplabel): os.mkdir( dir_pplabel ) pth_Label = os.path.join(dir_pplabel, 'Label.txt') Label_txt = '' Label_txt += f'''{dir_pplabel}/{m5}.jpg\t''' pth_img = os.path.join(dir, 'img/{}.txt'.format(m5)) pth_json = os.path.join(dir, 'json/{}.json'.format(m5)) if not os.path.exists(pth_img): # 没有相应的图片，可能被删除了 # raise Exception(f'Warnnig: no image {pth_img}') print( f'Warnnig: no image {pth_img}' ) return [] if not os.path.exists(pth_json): # raise Exception(f'Warnnig: no image {pth_json}') return [] if is_debug: with open(pth_img, "r", encoding="utf-8") as fp: imgdata = fp.read() imgdata = base64.b64decode(imgdata) imgdata = np.frombuffer(imgdata, np.uint8) img = cv2.imdecode(imgdata, cv2.IMREAD_UNCHANGED) # cv2.imshow('img', img) # cv2.waitKey(0) if len(img.shape) != 3: # 转彩图 img_color = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) # DBNet 原版只能处理彩图，这里转一下 else: img_color = img.copy() cv2.imwrite(os.path.join(dir_pplabel, f'{m5}.jpg'), img_color) jsn = load_json(pth_json) orgHeight = jsn['orgHeight'] orgWidth = jsn['orgWidth'] def left_right(jsn, img_color, Label_txt, pth_Label): """ 左右分栏模板 """ array = [] array2 = [] wordsInfo = jsn['prism_wordsInfo'] for j in range(len(wordsInfo)): jo = wordsInfo[j] word = jo["word"] # prism-wordsInfo 里的 angle 文字块的角度，这个角度只影响width和height，当角度为-90、90、-270、270，width和height的值需要自行互换 angle = jo['angle'] word_x = jo['x'] word_y = jo['y'] word_width = jo['width'] word_height = jo['height'] if abs(angle) == 90 or abs(angle) == 270: word_width = jo['height'] word_height = jo['width'] pos = jo['pos'] # 四个角的位置 左上、右上、右下、左下 顺时针方向 lu = [pos[0]['x'], pos[0]['y']] ru = [pos[1]['x'], pos[1]['y']] rd = [pos[2]['x'], pos[2]['y']] ld = [pos[3]['x'], pos[3]['y']] x1 = min( pos[0]['x'], pos[3]['x'] ) # 当前行的极左 x2 = max( pos[1]['x'], pos[2]['x'] ) # 当前行的极右 y1 = min( pos[0]['y'], pos[1]['y'] ) # 当前行的极上 y2 = max( pos[2]['y'], pos[3]['y'] ) # 当前行的极下 #img_color = cv2.rectangle(img_color, (x1, y1), (x2, y2), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 #img_color = cv2.rectangle(img_color, (lu[0], lu[1]), (rd[0], rd[1]), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) if match := re.compile('(^\d+)\.').search(word): nth = int( match.group(1) ) print( nth, word ) array.append( { "nth":nth, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) else: array2.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) pass transcriptions = [] last_x2 = 99999 x_right = -1 left = [] right = [] for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x_right == -1 and x1 > last_x2: # 第一次超过上一行的右边 x_right = x1 if x_right != -1: right.append( item ) else: left.append( item ) last_x2 = x2 if len(left) > 0 and len(right) > 0: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in left: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(left) > 0 and len(right) > 0: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in right: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) # [{"transcription":"","points":[[160,167],[1123,167],[1123,2970],[160,2970]]},{"transcription":"","points":[[1134,160],[2078,160],[2078,2933],[1134,2933]]}] # points 是框选的矩形四个角坐标： 左上 右上 右下 左下 transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(transcriptions) > 0: Label_txt += jsonstring( transcriptions ) Label_txt += '\n' with open(pth_Label, 'w', encoding='utf-8') as f: f.write(Label_txt) # huggingface/PPOCRLabel/PPOCRLabel.py 用这个打开目录 # ok 标注框是对的 cv2.imwrite(os.path.join(dir_pplabel, f'{m5}_auto_select.jpg'), img_color) return transcriptions def left_noright(jsn, img_color, Label_txt, pth_Label): """ 左右分栏模板 """ array = [] array2 = [] wordsInfo = jsn['prism_wordsInfo'] for j in range(len(wordsInfo)): jo = wordsInfo[j] word = jo["word"] # prism-wordsInfo 里的 angle 文字块的角度，这个角度只影响width和height，当角度为-90、90、-270、270，width和height的值需要自行互换 angle = jo['angle'] word_x = jo['x'] word_y = jo['y'] word_width = jo['width'] word_height = jo['height'] if abs(angle) == 90 or abs(angle) == 270: word_width = jo['height'] word_height = jo['width'] pos = jo['pos'] # 四个角的位置 左上、右上、右下、左下 顺时针方向 lu = [pos[0]['x'], pos[0]['y']] ru = [pos[1]['x'], pos[1]['y']] rd = [pos[2]['x'], pos[2]['y']] ld = [pos[3]['x'], pos[3]['y']] x1 = min( pos[0]['x'], pos[3]['x'] ) # 当前行的极左 x2 = max( pos[1]['x'], pos[2]['x'] ) # 当前行的极右 y1 = min( pos[0]['y'], pos[1]['y'] ) # 当前行的极上 y2 = max( pos[2]['y'], pos[3]['y'] ) # 当前行的极下 #img_color = cv2.rectangle(img_color, (x1, y1), (x2, y2), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 #img_color = cv2.rectangle(img_color, (lu[0], lu[1]), (rd[0], rd[1]), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) if match := re.compile('(^[\S]+)').search(word): g1 = match.group(1) # print( g1 ) array.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) array2.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) pass transcriptions = [] last_x2 = 99999 x_right = -1 left = [] right = [] for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x_right == -1 and x1 > last_x2: # 第一次超过上一行的右边 x_right = x1 if x_right != -1: right.append( item ) else: left.append( item ) last_x2 = x2 if len(left) > 5: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in left: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(right) > 5: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in right: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(transcriptions) > 0: Label_txt += jsonstring( transcriptions ) Label_txt += '\n' with open(pth_Label, 'w', encoding='utf-8') as f: f.write(Label_txt) # huggingface/PPOCRLabel/PPOCRLabel.py 用这个打开目录 # ok 标注框是对的 cv2.imwrite(os.path.join(dir_pplabel, f'{m5}_auto_select.jpg'), img_color) return transcriptions transcriptions = left_right(jsn, img_color, Label_txt, pth_Label) if len(transcriptions) > 0: return transcriptions transcriptions = left_noright(jsn, img_color, Label_txt, pth_Label) if len(transcriptions) > 0: return transcriptions return [] def get_selectionsv2(dir, m5): transcriptions = [] dir_pplabel = 'pplabel' if not os.path.exists(dir_pplabel): os.mkdir( dir_pplabel ) pth_Label = os.path.join(dir_pplabel, 'Label.txt') Label_txt = '' Label_txt += f'''{dir_pplabel}/{m5}.jpg\t''' pth_img = os.path.join(dir, 'img/{}.txt'.format(m5)) pth_json = os.path.join(dir, 'json/{}.json'.format(m5)) if not os.path.exists(pth_img): # 没有相应的图片，可能被删除了 # raise Exception(f'Warnnig: no image {pth_img}') print( f'Warnnig: no image {pth_img}' ) return [] if not os.path.exists(pth_json): # raise Exception(f'Warnnig: no image {pth_json}') print( f'Warnnig: no json {pth_json}' ) return [] img_color = None if is_debug: with open(pth_img, "r", encoding="utf-8") as fp: imgdata = fp.read() imgdata = base64.b64decode(imgdata) imgdata = np.frombuffer(imgdata, np.uint8) img = cv2.imdecode(imgdata, cv2.IMREAD_UNCHANGED) # cv2.imshow('img', img) # cv2.waitKey(0) if len(img.shape) != 3: # 转彩图 img_color = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) # DBNet 原版只能处理彩图，这里转一下 else: img_color = img.copy() cv2.imwrite(os.path.join(dir_pplabel, f'{m5}.jpg'), img_color) jsn = load_json(pth_json) orgHeight = jsn['orgHeight'] orgWidth = jsn['orgWidth'] def pos_info(jsn, img_color): array_digitNum = [] array_digitZH = [] array_ZH = [] array_other = [] wordsInfo = jsn['prism_wordsInfo'] for j in range(len(wordsInfo)): jo = wordsInfo[j] word = jo["word"] # prism-wordsInfo 里的 angle 文字块的角度，这个角度只影响width和height，当角度为-90、90、-270、270，width和height的值需要自行互换 angle = jo['angle'] word_x = jo['x'] word_y = jo['y'] word_width = jo['width'] word_height = jo['height'] if abs(angle) == 90 or abs(angle) == 270: word_width = jo['height'] word_height = jo['width'] pos = jo['pos'] # 四个角的位置 左上、右上、右下、左下 顺时针方向 lu = [pos[0]['x'], pos[0]['y']] ru = [pos[1]['x'], pos[1]['y']] rd = [pos[2]['x'], pos[2]['y']] ld = [pos[3]['x'], pos[3]['y']] x1 = min( pos[0]['x'], pos[3]['x'] ) # 当前行的极左 x2 = max( pos[1]['x'], pos[2]['x'] ) # 当前行的极右 y1 = min( pos[0]['y'], pos[1]['y'] ) # 当前行的极上 y2 = max( pos[2]['y'], pos[3]['y'] ) # 当前行的极下 #img_color = cv2.rectangle(img_color, (x1, y1), (x2, y2), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 #img_color = cv2.rectangle(img_color, (lu[0], lu[1]), (rd[0], rd[1]), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) if len(word) >= 10 and ( match := re.compile('(^\d+)\.[\S]').search(word) ): # if len(word) >= 10: nth = int( match.group(1) ) print('digitNum: ', nth) array_digitNum.append( { "nth":nth, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) elif match := re.compile('(^[一二三四五六七八九十]+、)').search(word): nth = match.group(1) print('digitZH: ', nth) array_digitZH.append( { "nth":nth, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) elif match := re.compile('(^[\u4e00-\u9fa5]+)').search(word): zh = match.group(1) print('ZH: ', zh) array_ZH.append( { "zh":zh, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) elif match := re.compile('(^[\S])').search(word): other = match.group(1) print('other: ', other) array_other.append( { "other":other, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) return array_digitNum, array_digitZH, array_ZH, array_other # 是否左右分栏 def is_left_right(array, orgWidth): def minmax(ar): """ 极左极右 """ min_x = 99999 max_x = -1 for item in ar: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x1 < min_x: min_x = x1 if x2 > max_x: max_x = x2 return min_x, max_x def relatedWidth(min_x, max_x): """ 相对宽度 """ w = max_x - min_x return w min_x, max_x = minmax(array) rltWidth = relatedWidth( min_x, max_x ) last_x2 = 99999 x_right = -1 left = [] right = [] flag = False for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if not flag and x1 > rltWidth / 2: right.append( item ) continue elif x1 <= rltWidth / 2: flag = True if x_right == -1 and x1 > last_x2: # 第一次超过上一行的右边 x_right = x1 if x_right != -1 and x1 > last_x2: right.append( item ) else: left.append( item ) if x_right == -1: last_x2 = x2 # if len(left) > 0 and len(right) >0: # return True # return False return left, right def middle_left_right(left, right): """ 找出左右分栏中间区域的极左和极右 """ # 左边找极右 max_right = -1 for idx, item in enumerate(left): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x2 > max_right: max_right = x2 # 右边找极左 min_left = 99999 for idx, item in enumerate(right): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x1 < min_left: min_left = x1 return min_left, max_right def left_or_right( min_left, max_right, array ): """ max_right: 左边极右（左右分栏的左栏） min_left: 右边极左（左右分栏的右栏） """ left = [] right = [] for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x1 < min_left and x2 < min_left: # 它的左边小于的右边极左，并且它的右边也小于的右边极左 left.append( item ) elif x1 > max_right: # 它的左边大于中间区域的右边 right.append( item ) else: print(f'Waring: not left AND not right. {item}') return left, right def get_selection(array, img_color): transcriptions = [] if len(array) > 0: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in array: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if is_debug: img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 resized_image = show_img(img_color) cv2.imwrite(os.path.join(dir_pplabel, f'{m5}_select.jpg'), resized_image) # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) return transcriptions def left_right(jsn, img_color, Label_txt, pth_Label): """ 左右分栏模板 """ array = [] array2 = [] wordsInfo = jsn['prism_wordsInfo'] for j in range(len(wordsInfo)): jo = wordsInfo[j] word = jo["word"] # prism-wordsInfo 里的 angle 文字块的角度，这个角度只影响width和height，当角度为-90、90、-270、270，width和height的值需要自行互换 angle = jo['angle'] word_x = jo['x'] word_y = jo['y'] word_width = jo['width'] word_height = jo['height'] if abs(angle) == 90 or abs(angle) == 270: word_width = jo['height'] word_height = jo['width'] pos = jo['pos'] # 四个角的位置 左上、右上、右下、左下 顺时针方向 lu = [pos[0]['x'], pos[0]['y']] ru = [pos[1]['x'], pos[1]['y']] rd = [pos[2]['x'], pos[2]['y']] ld = [pos[3]['x'], pos[3]['y']] x1 = min( pos[0]['x'], pos[3]['x'] ) # 当前行的极左 x2 = max( pos[1]['x'], pos[2]['x'] ) # 当前行的极右 y1 = min( pos[0]['y'], pos[1]['y'] ) # 当前行的极上 y2 = max( pos[2]['y'], pos[3]['y'] ) # 当前行的极下 #img_color = cv2.rectangle(img_color, (x1, y1), (x2, y2), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 #img_color = cv2.rectangle(img_color, (lu[0], lu[1]), (rd[0], rd[1]), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) if match := re.compile('(^\d+)\.').search(word): nth = int( match.group(1) ) print( nth, word ) array.append( { "nth":nth, "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) else: array2.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) pass transcriptions = [] last_x2 = 99999 x_right = -1 left = [] right = [] for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x_right == -1 and x1 > last_x2: # 第一次超过上一行的右边 x_right = x1 if x_right != -1: right.append( item ) else: left.append( item ) last_x2 = x2 if len(left) > 0 and len(right) > 0: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in left: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(left) > 0 and len(right) > 0: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in right: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) # [{"transcription":"","points":[[160,167],[1123,167],[1123,2970],[160,2970]]},{"transcription":"","points":[[1134,160],[2078,160],[2078,2933],[1134,2933]]}] # points 是框选的矩形四个角坐标： 左上 右上 右下 左下 transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(transcriptions) > 0: Label_txt += jsonstring( transcriptions ) Label_txt += '\n' with open(pth_Label, 'w', encoding='utf-8') as f: f.write(Label_txt) # huggingface/PPOCRLabel/PPOCRLabel.py 用这个打开目录 # ok 标注框是对的 cv2.imwrite(os.path.join(dir_pplabel, f'{m5}_auto_select.jpg'), img_color) return transcriptions def left_noright(jsn, img_color, Label_txt, pth_Label): """ 左右分栏模板 """ array = [] array2 = [] wordsInfo = jsn['prism_wordsInfo'] for j in range(len(wordsInfo)): jo = wordsInfo[j] word = jo["word"] # prism-wordsInfo 里的 angle 文字块的角度，这个角度只影响width和height，当角度为-90、90、-270、270，width和height的值需要自行互换 angle = jo['angle'] word_x = jo['x'] word_y = jo['y'] word_width = jo['width'] word_height = jo['height'] if abs(angle) == 90 or abs(angle) == 270: word_width = jo['height'] word_height = jo['width'] pos = jo['pos'] # 四个角的位置 左上、右上、右下、左下 顺时针方向 lu = [pos[0]['x'], pos[0]['y']] ru = [pos[1]['x'], pos[1]['y']] rd = [pos[2]['x'], pos[2]['y']] ld = [pos[3]['x'], pos[3]['y']] x1 = min( pos[0]['x'], pos[3]['x'] ) # 当前行的极左 x2 = max( pos[1]['x'], pos[2]['x'] ) # 当前行的极右 y1 = min( pos[0]['y'], pos[1]['y'] ) # 当前行的极上 y2 = max( pos[2]['y'], pos[3]['y'] ) # 当前行的极下 #img_color = cv2.rectangle(img_color, (x1, y1), (x2, y2), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 #img_color = cv2.rectangle(img_color, (lu[0], lu[1]), (rd[0], rd[1]), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) if match := re.compile('(^[\S]+)').search(word): g1 = match.group(1) # print( g1 ) array.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) array2.append( { "line":word, "leftup":(x1, y1), "rightdown":(x2, y2) } ) pass transcriptions = [] last_x2 = 99999 x_right = -1 left = [] right = [] for idx, item in enumerate(array): leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] if x_right == -1 and x1 > last_x2: # 第一次超过上一行的右边 x_right = x1 if x_right != -1: right.append( item ) else: left.append( item ) last_x2 = x2 if len(left) > 5: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in left: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(right) > 5: min_x1_left = 99999 min_y1_left = 99999 max_x2_left = -1 max_y2_left = -1 for item in right: leftup = item['leftup'] rightdown = item['rightdown'] x1 = leftup[0] x2 = rightdown[0] y1 = leftup[1] y2 = rightdown[1] if x1 < min_x1_left: min_x1_left = x1 if y1 < min_y1_left: min_y1_left = y1 if x2 > max_x2_left: max_x2_left = x2 if y2 > max_y2_left: max_y2_left = y2 if len(array2)> 0: for item in array2: if item['leftup'][1] < min_y1_left: min_y1_left = item['leftup'][1] if item['rightdown'][1] > max_y2_left: max_y2_left = item['rightdown'][1] img_color = cv2.rectangle(img_color, (min_x1_left, min_y1_left), (max_x2_left, max_y2_left), (0, 255, 0), 2) # 矩形的左上角, 矩形的右下角 # cv2.imshow("green", img_color) # cv2.waitKey(0) transcriptions.append( {"transcription": "", "points":[ [min_x1_left, min_y1_left], [max_x2_left, min_y1_left], [max_x2_left, max_y2_left], [min_x1_left, max_y2_left] ]} ) if len(transcriptions) > 0: Label_txt += jsonstring( transcriptions ) Label_txt += '\n' with open(pth_Label, 'w', encoding='utf-8') as f: f.write(Label_txt) # huggingface/PPOCRLabel/PPOCRLabel.py 用这个打开目录 # ok 标注框是对的 cv2.imwrite(os.path.join(dir_pplabel, f'{m5}_auto_select.jpg'), img_color) return transcriptions array_digitNum, array_digitZH, array_ZH, array_other = pos_info(jsn, img_color) left_digitNum, right_digitNum = is_left_right(array_digitNum, orgWidth) left_digitZH, right_digitZH = is_left_right(array_digitZH, orgWidth) if len(left_digitNum) > 0 and len(right_digitNum) > 0: """ 左右分栏 """ # 找中间区域的左右 x 坐标 我 min_left, max_right = middle_left_right(left_digitNum, right_digitNum) # 乘下的东西全部分成左右两边 left, right = left_or_right( min_left, max_right, array_digitZH + array_ZH + array_other ) left_filted = [ item for item in left if len(item['line']) > 1 ] right_filted = [ item for item in right if len(item['line']) > 1 ] transcriptions_left = get_selection(left_digitNum + left_filted, img_color) transcriptions_right = get_selection(right_digitNum + right_filted, img_color) transcriptions = transcriptions_left + transcriptions_right pass elif len(left_digitZH) > 0 and len(right_digitZH) > 0: # 左右分栏 pass else: # 单栏 transcriptions = get_selection(array_digitNum + array_digitZH + array_ZH + array_other, img_color) # transcriptions = left_right(jsn, img_color, Label_txt, pth_Label) # if len(transcriptions) > 0: # return transcriptions # transcriptions = left_noright(jsn, img_color, Label_txt, pth_Label) # if len(transcriptions) > 0: # return transcriptions return transcriptions @app.route('/img/autoselection', methods=['post']) def autoselection(): global dic_cache # request.json 只能够接受方法为POST、Body为raw，header 内容为 application/json类型的数据 print(request.json, type(request.json)) dir = '/yingedu/project/ocr_server_test/data/' import sys if sys.platform.startswith('win'): dir = 'D:/workcode/nodejs/ocr/tools/data' m5 = request.json['m5'] if m5 in dic_cache: transcriptions = dic_cache[m5] print( '########## hint cache.' ) else: transcriptions = get_selectionsv2(dir, m5) dic_cache[m5] = transcriptions # print( transcriptions ) # ret = { "status": 200, "data":transcriptions } return jsonify(transcriptions) if __name__ == '__main__': dir = '/yingedu/project/ocr_server_test/data/' m5 = 'b0cd8663d03d596eab297a8250bf116c' mds = [ 'b0cd8663d03d596eab297a8250bf116c', '002a5d44e1771320fdcddb5df4e3cdb1', '3ce51feeb9a548e373fe72eb28df4a6b', '4b890c40319f8857be7a95705d8e61a3', 'b4ea47ca3af4c2862b75299fdaf103a8', 'd153c1c67a9502584c579ea7dd8f54e5', '9c80bd010a27cc410e77eea76170c829', 'ab35bfe44d130564efce70bf5f1152cd', '794629e25384df1b8b49936a7159166d', 'b89548a9c03b2de2d19c10e9173d9624', '756819dd53a98691859de016f7138d71', '6f63f59e6b826fa8a3284282cb503574', '5f232626c6d37267733c4383bea0550f', '5269bc6508598a8ee5dbd753ea353a3c', ] # mds = [ # '020bbac5534add85eb9d5bce4afedfb3', # ] if is_debug: dir = 'D:/workcode/nodejs/ocr/tools/data' for m in mds: transcriptions = get_selectionsv2(dir, m) print( transcriptions ) else: app.run(host="0.0.0.0", port=2393, debug=True)