PaddleOCR/tools/infer/utility.py

# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import argparse
import os, sys
from ppocr.utils.utility import initial_logger
logger = initial_logger()
from paddle.fluid.core import PaddleTensor
from paddle.fluid.core import AnalysisConfig
from paddle.fluid.core import create_paddle_predictor
import cv2
import numpy as np
import json
from PIL import Image, ImageDraw, ImageFont
import math


def parse_args():
    def str2bool(v):
        return v.lower() in ("true", "t", "1")

    parser = argparse.ArgumentParser()
    #params for prediction engine
    parser.add_argument("--use_gpu", type=str2bool, default=True)
    parser.add_argument("--ir_optim", type=str2bool, default=True)
    parser.add_argument("--use_tensorrt", type=str2bool, default=False)
    parser.add_argument("--gpu_mem", type=int, default=8000)

    #params for text detector
    parser.add_argument("--image_dir", type=str)
    parser.add_argument("--det_algorithm", type=str, default='DB')
    parser.add_argument("--det_model_dir", type=str)
    parser.add_argument("--det_max_side_len", type=float, default=960)

    #DB parmas
    parser.add_argument("--det_db_thresh", type=float, default=0.3)
    parser.add_argument("--det_db_box_thresh", type=float, default=0.5)
    parser.add_argument("--det_db_unclip_ratio", type=float, default=2.0)

    #EAST parmas
    parser.add_argument("--det_east_score_thresh", type=float, default=0.8)
    parser.add_argument("--det_east_cover_thresh", type=float, default=0.1)
    parser.add_argument("--det_east_nms_thresh", type=float, default=0.2)

    #params for text recognizer
    parser.add_argument("--rec_algorithm", type=str, default='CRNN')
    parser.add_argument("--rec_model_dir", type=str)
    parser.add_argument("--rec_image_shape", type=str, default="3, 32, 320")
    parser.add_argument("--rec_char_type", type=str, default='ch')
    parser.add_argument("--rec_batch_num", type=int, default=30)
    parser.add_argument(
        "--rec_char_dict_path",
        type=str,
        default="./ppocr/utils/ppocr_keys_v1.txt")
    parser.add_argument("--use_space_char", type=bool, default=True)
    return parser.parse_args()


def create_predictor(args, mode):
    if mode == "det":
        model_dir = args.det_model_dir
    else:
        model_dir = args.rec_model_dir

    if model_dir is None:
        logger.info("not find {} model file path {}".format(mode, model_dir))
        sys.exit(0)
    model_file_path = model_dir + "/model"
    params_file_path = model_dir + "/params"
    if not os.path.exists(model_file_path):
        logger.info("not find model file path {}".format(model_file_path))
        sys.exit(0)
    if not os.path.exists(params_file_path):
        logger.info("not find params file path {}".format(params_file_path))
        sys.exit(0)

    config = AnalysisConfig(model_file_path, params_file_path)

    if args.use_gpu:
        config.enable_use_gpu(args.gpu_mem, 0)
    else:
        config.disable_gpu()
        # config.enable_mkldnn()
        config.set_cpu_math_library_num_threads(4)
    #config.enable_memory_optim()
    config.disable_glog_info()

    # use zero copy
    config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass")
    config.switch_use_feed_fetch_ops(False)
    predictor = create_paddle_predictor(config)
    input_names = predictor.get_input_names()
    input_tensor = predictor.get_input_tensor(input_names[0])
    output_names = predictor.get_output_names()
    output_tensors = []
    for output_name in output_names:
        output_tensor = predictor.get_output_tensor(output_name)
        output_tensors.append(output_tensor)
    return predictor, input_tensor, output_tensors


def draw_text_det_res(dt_boxes, img_path):
    src_im = cv2.imread(img_path)
    for box in dt_boxes:
        box = np.array(box).astype(np.int32).reshape(-1, 2)
        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
    return src_im


def resize_img(img, input_size=600):
    """
    resize img and limit the longest side of the image to input_size
    """
    img = np.array(img)
    im_shape = img.shape
    im_size_max = np.max(im_shape[0:2])
    im_scale = float(input_size) / float(im_size_max)
    im = cv2.resize(img, None, None, fx=im_scale, fy=im_scale)
    return im


def draw_ocr(image, boxes, txts, scores, draw_txt=True, drop_score=0.5):
    """
    Visualize the results of OCR detection and recognition
    args:
        image(Image|array): RGB image
        boxes(list): boxes with shape(N, 4, 2)
        txts(list): the texts
        scores(list): txxs corresponding scores
        draw_txt(bool): whether draw text or not
        drop_score(float): only scores greater than drop_threshold will be visualized
    return(array):
        the visualized img
    """
    if scores is None:
        scores = [1] * len(boxes)
    for (box, score) in zip(boxes, scores):
        if score < drop_score or math.isnan(score):
            continue
        box = np.reshape(np.array(box), [-1, 1, 2]).astype(np.int64)
        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)

    if draw_txt:
        img = np.array(resize_img(image, input_size=600))
        txt_img = text_visual(
            txts, scores, img_h=img.shape[0], img_w=600, threshold=drop_score)
        img = np.concatenate([np.array(img), np.array(txt_img)], axis=1)
        return img
    return image


def draw_ocr_box_txt(image, boxes, txts):
    h, w = image.height, image.width
    img_left = image.copy()
    img_right = Image.new('RGB', (w, h), (255, 255, 255))

    import random
    # 每次使用相同的随机种子 ，可以保证两次颜色一致
    random.seed(0)
    draw_left = ImageDraw.Draw(img_left)
    draw_right = ImageDraw.Draw(img_right)
    for (box, txt) in zip(boxes, txts):
        color = (random.randint(0, 255), random.randint(0, 255),
                 random.randint(0, 255))
        draw_left.polygon(box, fill=color)
        draw_right.polygon(
            [
                box[0][0], box[0][1], box[1][0], box[1][1], box[2][0],
                box[2][1], box[3][0], box[3][1]
            ],
            outline=color)
        box_height = math.sqrt((box[0][0] - box[3][0])**2 + (box[0][1] - box[3][
            1])**2)
        box_width = math.sqrt((box[0][0] - box[1][0])**2 + (box[0][1] - box[1][
            1])**2)
        if box_height > 2 * box_width:
            font_size = max(int(box_width * 0.9), 10)
            font = ImageFont.truetype(
                "./doc/simfang.ttf", font_size, encoding="utf-8")
            cur_y = box[0][1]
            for c in txt:
                char_size = font.getsize(c)
                draw_right.text(
                    (box[0][0] + 3, cur_y), c, fill=(0, 0, 0), font=font)
                cur_y += char_size[1]
        else:
            font_size = max(int(box_height * 0.8), 10)
            font = ImageFont.truetype(
                "./doc/simfang.ttf", font_size, encoding="utf-8")
            draw_right.text(
                [box[0][0], box[0][1]], txt, fill=(0, 0, 0), font=font)
    img_left = Image.blend(image, img_left, 0.5)
    img_show = Image.new('RGB', (w * 2, h), (255, 255, 255))
    img_show.paste(img_left, (0, 0, w, h))
    img_show.paste(img_right, (w, 0, w * 2, h))
    return np.array(img_show)


def str_count(s):
    """
    Count the number of Chinese characters,
    a single English character and a single number
    equal to half the length of Chinese characters.

    args:
        s(string): the input of string
    return(int):
        the number of Chinese characters
    """
    import string
    count_zh = count_pu = 0
    s_len = len(s)
    en_dg_count = 0
    for c in s:
        if c in string.ascii_letters or c.isdigit() or c.isspace():
            en_dg_count += 1
        elif c.isalpha():
            count_zh += 1
        else:
            count_pu += 1
    return s_len - math.ceil(en_dg_count / 2)


def text_visual(texts, scores, img_h=400, img_w=600, threshold=0.):
    """
    create new blank img and draw txt on it
    args:
        texts(list): the text will be draw
        scores(list|None): corresponding score of each txt
        img_h(int): the height of blank img
        img_w(int): the width of blank img
    return(array):

    """
    if scores is not None:
        assert len(texts) == len(
            scores), "The number of txts and corresponding scores must match"

    def create_blank_img():
        blank_img = np.ones(shape=[img_h, img_w], dtype=np.int8) * 255
        blank_img[:, img_w - 1:] = 0
        blank_img = Image.fromarray(blank_img).convert("RGB")
        draw_txt = ImageDraw.Draw(blank_img)
        return blank_img, draw_txt

    blank_img, draw_txt = create_blank_img()

    font_size = 20
    txt_color = (0, 0, 0)
    font = ImageFont.truetype("./doc/simfang.ttf", font_size, encoding="utf-8")

    gap = font_size + 5
    txt_img_list = []
    count, index = 1, 0
    for idx, txt in enumerate(texts):
        index += 1
        if scores[idx] < threshold or math.isnan(scores[idx]):
            index -= 1
            continue
        first_line = True
        while str_count(txt) >= img_w // font_size - 4:
            tmp = txt
            txt = tmp[:img_w // font_size - 4]
            if first_line:
                new_txt = str(index) + ': ' + txt
                first_line = False
            else:
                new_txt = '    ' + txt
            draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
            txt = tmp[img_w // font_size - 4:]
            if count >= img_h // gap - 1:
                txt_img_list.append(np.array(blank_img))
                blank_img, draw_txt = create_blank_img()
                count = 0
            count += 1
        if first_line:
            new_txt = str(index) + ': ' + txt + '   ' + '%.3f' % (scores[idx])
        else:
            new_txt = "  " + txt + "  " + '%.3f' % (scores[idx])
        draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
        # whether add new blank img or not
        if count >= img_h // gap - 1 and idx + 1 < len(texts):
            txt_img_list.append(np.array(blank_img))
            blank_img, draw_txt = create_blank_img()
            count = 0
        count += 1
    txt_img_list.append(np.array(blank_img))
    if len(txt_img_list) == 1:
        blank_img = np.array(txt_img_list[0])
    else:
        blank_img = np.concatenate(txt_img_list, axis=1)
    return np.array(blank_img)


def base64_to_cv2(b64str):
    import base64
    data = base64.b64decode(b64str.encode('utf8'))
    data = np.fromstring(data, np.uint8)
    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
    return data


def draw_boxes(image, boxes, scores=None, drop_score=0.5):
    if scores is None:
        scores = [1] * len(boxes)
    for (box, score) in zip(boxes, scores):
        if score < drop_score:
            continue
        box = np.reshape(np.array(box), [-1, 1, 2]).astype(np.int64)
        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
    return image


if __name__ == '__main__':
    test_img = "./doc/test_v2"
    predict_txt = "./doc/predict.txt"
    f = open(predict_txt, 'r')
    data = f.readlines()
    img_path, anno = data[0].strip().split('\t')
    img_name = os.path.basename(img_path)
    img_path = os.path.join(test_img, img_name)
    image = Image.open(img_path)

    data = json.loads(anno)
    boxes, txts, scores = [], [], []
    for dic in data:
        boxes.append(dic['points'])
        txts.append(dic['transcription'])
        scores.append(round(dic['scores'], 3))

    new_img = draw_ocr(image, boxes, txts, scores, draw_txt=True)

    cv2.imwrite(img_name, new_img)