predict_rec.py 4.9 KB
Newer Older
L
LDOUBLEV 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
# 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 utility
from ppocr.utils.utility import initial_logger
logger = initial_logger()
import cv2

import copy
import numpy as np
import math
import time
from ppocr.utils.character import CharacterOps


class TextRecognizer(object):
    def __init__(self, args):
        self.predictor, self.input_tensor, self.output_tensors =\
            utility.create_predictor(args, mode="rec")
        image_shape = [int(v) for v in args.rec_image_shape.split(",")]
        self.rec_image_shape = image_shape
        char_ops_params = {}
        char_ops_params["character_type"] = args.rec_char_type
        char_ops_params["character_dict_path"] = args.rec_char_dict_path
        char_ops_params['loss_type'] = 'ctc'
        self.char_ops = CharacterOps(char_ops_params)

39
    def resize_norm_img(self, img, max_wh_ratio):
L
LDOUBLEV 已提交
40
        imgC, imgH, imgW = self.rec_image_shape
41
        imgW = int(32 * max_wh_ratio)
L
LDOUBLEV 已提交
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
        h = img.shape[0]
        w = img.shape[1]
        ratio = w / float(h)
        if math.ceil(imgH * ratio) > imgW:
            resized_w = imgW
        else:
            resized_w = int(math.ceil(imgH * ratio))
        resized_image = cv2.resize(img, (resized_w, imgH))
        resized_image = resized_image.astype('float32')
        resized_image = resized_image.transpose((2, 0, 1)) / 255
        resized_image -= 0.5
        resized_image /= 0.5
        padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32)
        padding_im[:, :, 0:resized_w] = resized_image
        return padding_im

    def __call__(self, img_list):
        img_num = len(img_list)
60
        batch_num = 30
L
LDOUBLEV 已提交
61 62 63 64 65
        rec_res = []
        predict_time = 0
        for beg_img_no in range(0, img_num, batch_num):
            end_img_no = min(img_num, beg_img_no + batch_num)
            norm_img_batch = []
66
            max_wh_ratio = 0
L
LDOUBLEV 已提交
67
            for ino in range(beg_img_no, end_img_no):
68 69 70 71 72
                h, w = img_list[ino].shape[0:2]
                wh_ratio = w * 1.0 / h
                max_wh_ratio = max(max_wh_ratio, wh_ratio)
            for ino in range(beg_img_no, end_img_no):
                norm_img = self.resize_norm_img(img_list[ino], max_wh_ratio)
L
LDOUBLEV 已提交
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
                norm_img = norm_img[np.newaxis, :]
                norm_img_batch.append(norm_img)
            norm_img_batch = np.concatenate(norm_img_batch)
            norm_img_batch = norm_img_batch.copy()
            starttime = time.time()
            self.input_tensor.copy_from_cpu(norm_img_batch)
            self.predictor.zero_copy_run()
            rec_idx_batch = self.output_tensors[0].copy_to_cpu()
            rec_idx_lod = self.output_tensors[0].lod()[0]
            predict_batch = self.output_tensors[1].copy_to_cpu()
            predict_lod = self.output_tensors[1].lod()[0]
            elapse = time.time() - starttime
            predict_time += elapse
            starttime = time.time()
            for rno in range(len(rec_idx_lod) - 1):
                beg = rec_idx_lod[rno]
                end = rec_idx_lod[rno + 1]
                rec_idx_tmp = rec_idx_batch[beg:end, 0]
                preds_text = self.char_ops.decode(rec_idx_tmp)
                beg = predict_lod[rno]
                end = predict_lod[rno + 1]
                probs = predict_batch[beg:end, :]
                ind = np.argmax(probs, axis=1)
                blank = probs.shape[1]
                valid_ind = np.where(ind != (blank - 1))[0]
                score = np.mean(probs[valid_ind, ind[valid_ind]])
                rec_res.append([preds_text, score])
        return rec_res, predict_time


if __name__ == "__main__":
    args = utility.parse_args()
    image_file_list = utility.get_image_file_list(args.image_dir)
    text_recognizer = TextRecognizer(args)
    valid_image_file_list = []
    img_list = []
    for image_file in image_file_list:
        img = cv2.imread(image_file)
        if img is None:
            logger.info("error in loading image:{}".format(image_file))
            continue
        valid_image_file_list.append(image_file)
        img_list.append(img)
    rec_res, predict_time = text_recognizer(img_list)
    rec_res, predict_time = text_recognizer(img_list)
    for ino in range(len(img_list)):
        print("Predicts of %s:%s" % (valid_image_file_list[ino], rec_res[ino]))
    print("Total predict time for %d images:%.3f" %
          (len(img_list), predict_time))