utils.py

# -*- coding:utf-8 -*-
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import base64
import math

import cv2
import numpy as np
from PIL import Image
from PIL import ImageDraw
from PIL import ImageFont


def draw_ocr(image, boxes, txts, scores, font_file, 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, font_file, 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 text_visual(texts, scores, font_file, 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(font_file, 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 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 resize_img(img, input_size=600):
    img = np.array(img)
    im_shape = img.shape
    im_size_min = np.min(im_shape[0:2])
    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 get_image_ext(image):
    if image.shape[2] == 4:
        return ".png"
    return ".jpg"


def sorted_boxes(dt_boxes):
    """
    Sort text boxes in order from top to bottom, left to right
    args:
        dt_boxes(array):detected text boxes with shape [4, 2]
    return:
        sorted boxes(array) with shape [4, 2]
    """
    num_boxes = dt_boxes.shape[0]
    sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0]))
    _boxes = list(sorted_boxes)

    for i in range(num_boxes - 1):
        if abs(_boxes[i + 1][0][1] - _boxes[i][0][1]) < 10 and \
                (_boxes[i + 1][0][0] < _boxes[i][0][0]):
            tmp = _boxes[i]
            _boxes[i] = _boxes[i + 1]
            _boxes[i + 1] = tmp
    return _boxes


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