Merge pull request #754 from tink2123/lan_doc

multi-languages doc

configs/rec/multi_languages/rec_en_lite_train.yml

@@ -16,7 +16,7 @@ Global:
   loss_type: ctc
   distort: false
   use_space_char: false
-  reader_yml: ./configs/rec/rec_en_reader.yml
+  reader_yml: ./configs/rec/multi_languages/rec_en_reader.yml
   pretrain_weights:
   checkpoints:
   save_inference_dir:

configs/rec/multi_languages/rec_french_lite_train.yml

@@ -16,7 +16,7 @@ Global:
   loss_type: ctc
   distort: true
   use_space_char: false
-  reader_yml: ./configs/rec/rec_french_reader.yml
+  reader_yml: ./configs/rec/multi_languages/rec_french_reader.yml
   pretrain_weights:
   checkpoints:
   save_inference_dir:

configs/rec/multi_languages/rec_ger_lite_train.yml

@@ -16,7 +16,7 @@ Global:
   loss_type: ctc
   distort: true
   use_space_char: false
-  reader_yml: ./configs/rec/rec_ger_reader.yml
+  reader_yml: ./configs/rec/multi_languages/rec_ger_reader.yml
   pretrain_weights:
   checkpoints:
   save_inference_dir:

configs/rec/multi_languages/rec_japan_lite_train.yml

@@ -16,7 +16,7 @@ Global:
   loss_type: ctc
   distort: true
   use_space_char: false
-  reader_yml: ./configs/rec/rec_japan_reader.yml
+  reader_yml: ./configs/rec/multi_languages/rec_japan_reader.yml
   pretrain_weights:
   checkpoints:
   save_inference_dir:

configs/rec/multi_languages/rec_korean_lite_train.yml

@@ -16,7 +16,7 @@ Global:
   loss_type: ctc
   distort: true
   use_space_char: false
-  reader_yml: ./configs/rec/rec_korean_reader.yml
+  reader_yml: ./configs/rec/multi_languages/rec_korean_reader.yml
   pretrain_weights:
   checkpoints:
   save_inference_dir:

doc/doc_ch/inference.md

@@ -24,6 +24,7 @@ inference 模型（`fluid.io.save_inference_model`保存的模型）
     - [2. 基于CTC损失的识别模型推理](#基于CTC损失的识别模型推理)
     - [3. 基于Attention损失的识别模型推理](#基于Attention损失的识别模型推理)
     - [4. 自定义文本识别字典的推理](#自定义文本识别字典的推理)
+    - [5. 多语言模型的推理](#多语言模型的推理)
 
 - [四、方向分类模型推理](#方向识别模型推理)
     - [1. 方向分类模型推理](#方向分类模型推理)
@@ -305,6 +306,22 @@ dict_character = list(self.character_str)
 python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_char_dict_path="your text dict path"
 ```
 
+<a name="多语言模型的推理"></a>
+### 5. 多语言模型的推理
+如果您需要预测的是其他语言模型，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径, 同时为了得到正确的可视化结果，
+需要通过 `--vis_font_path` 指定可视化的字体路径，`doc/` 路径下有默认提供的小语种字体，例如韩文识别：
+
+```
+python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/korean/1.jpg" --rec_model_dir="./your inference model" --rec_char_type="korean" --rec_char_dict_path="ppocr/utils/korean_dict.txt" --vis_font_path="doc/korean.ttf"
+```
+![](../imgs_words/korean/1.jpg)
+
+执行命令后，上图的预测结果为：
+``` text
+2020-09-19 16:15:05,076-INFO: 	 index: [205 206  38  39]
+2020-09-19 16:15:05,077-INFO: 	 word : 바탕으로
+2020-09-19 16:15:05,077-INFO: 	 score: 0.9171358942985535
+```
 
 <a name="方向分类模型推理"></a>
 ## 四、方向分类模型推理

doc/doc_ch/models_list.md

@@ -49,10 +49,11 @@ PaddleOCR提供的可下载模型包括`预测模型`、`训练模型`、`预训
 #### 3. 多语言识别模型（更多语言持续更新中...）
 |模型名称|模型简介|预测模型大小|下载地址|
 |-|-|-|-|
-|-|法文识别|-|[预测模型]() / [训练模型]()|
-|-|德文识别|-|[预测模型]() / [训练模型]()|
-|-|韩文识别|-|[预测模型]() / [训练模型]()|
-|-|日文识别|-|[预测模型]() / [训练模型]()|
+| french_ppocr_mobile_v1.1_rec |法文识别|2.1M|[预测模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_train.tar)|
+| german_ppocr_mobile_v1.1_rec |德文识别|2.1M|[预测模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_train.tar)|
+| korean_ppocr_mobile_v1.1_rec |韩文识别|3.4M|[预测模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_train.tar)|
+| japan_ppocr_mobile_v1.1_rec |日文识别|3.7M|[预测模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_train.tar)|
+
 
 <a name="文本方向分类模型"></a>
 ### 三、文本方向分类模型

doc/doc_ch/recognition.md

@@ -208,7 +208,19 @@ Optimizer:
 ```
 **注意，预测/评估时的配置文件请务必与训练一致。**
 
+- 小语种
 
+PaddleOCR也提供了多语言的， `configs/rec/multi_languages` 路径下的提供了多语言的配置文件，目前PaddleOCR支持的多语言算法有：
+
+| 配置文件 |  算法名称 |   backbone |   trans   |   seq      |     pred     |  language ｜
+| :--------: |  :-------:   | :-------:  |   :-------:   |   :-----:   |  :-----:   | :-----:   |
+| rec_en_lite_train.yml |  CRNN |   Mobilenet_v3 small 0.5 |  None   |  BiLSTM |  ctc  | 英语   ｜
+| rec_french_lite_train.yml |  CRNN |   Mobilenet_v3 small 0.5 |  None   |  BiLSTM |  ctc  | 法语   ｜
+| rec_ger_lite_train.yml |  CRNN |   Mobilenet_v3 small 0.5 |  None   |  BiLSTM |  ctc  | 德语   ｜
+| rec_japan_lite_train.yml |  CRNN |   Mobilenet_v3 small 0.5 |  None   |  BiLSTM |  ctc  | 日语   ｜
+| rec_korean_lite_train.yml |  CRNN |   Mobilenet_v3 small 0.5 |  None   |  BiLSTM |  ctc  | 韩语   ｜
+
+多语言模型训练方式与中文模型一致，训练数据集均为100w的合成数据，少量的字体和测试数据可以在[百度网盘]()上下载。
 
 ### 评估
 

tools/infer/predict_system.py

@@ -133,6 +133,7 @@ def main(args):
     image_file_list = get_image_file_list(args.image_dir)
     text_sys = TextSystem(args)
     is_visualize = True
+    font_path = args.vis_font_path
     for image_file in image_file_list:
         img, flag = check_and_read_gif(image_file)
         if not flag:
@@ -160,7 +161,7 @@ def main(args):
             scores = [rec_res[i][1] for i in range(len(rec_res))]
 
             draw_img = draw_ocr(
-                image, boxes, txts, scores, drop_score=drop_score)
+                image, boxes, txts, scores, drop_score=drop_score, font_path=font_path)
             draw_img_save = "./inference_results/"
             if not os.path.exists(draw_img_save):
                 os.makedirs(draw_img_save)

tools/infer/utility.py

@@ -71,6 +71,10 @@ def parse_args():
         type=str,
         default="./ppocr/utils/ppocr_keys_v1.txt")
     parser.add_argument("--use_space_char", type=str2bool, default=True)
+    parser.add_argument(
+        "--vis_font_path",
+        type=str,
+        default="./doc/simfang.ttf")
 
     # params for text classifier
     parser.add_argument("--use_angle_cls", type=str2bool, default=False)
@@ -199,7 +203,7 @@ def draw_ocr(image,
     return image
 
 
-def draw_ocr_box_txt(image, boxes, txts):
+def draw_ocr_box_txt(image, boxes, txts, font_path="./doc/simfang.ttf"):
     h, w = image.height, image.width
     img_left = image.copy()
     img_right = Image.new('RGB', (w, h), (255, 255, 255))
@@ -226,7 +230,7 @@ def draw_ocr_box_txt(image, boxes, txts):
         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")
+                font_path, font_size, encoding="utf-8")
             cur_y = box[0][1]
             for c in txt:
                 char_size = font.getsize(c)
@@ -236,7 +240,7 @@ def draw_ocr_box_txt(image, boxes, txts):
         else:
             font_size = max(int(box_height * 0.8), 10)
             font = ImageFont.truetype(
-                "./doc/simfang.ttf", font_size, encoding="utf-8")
+                font_path, 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)