Skip to content

📄 Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.

修改模型转换说明Readme,增加26国语言rec模型,对应修改tools下的转换脚本。

benjaminwanfc3918ff519d
diff --git a/models/readme.md b/models/readme.md
index e38500c..e609921 100755
--- a/models/readme.md
+++ b/models/readme.md
@@ -1,93 +1,302 @@
## Download models
+
* from baiduNetDisk:
-链接:https://pan.baidu.com/s/1UhDrC7iOMWiYaQiHk3acdw 提取码:yjgv
+链接:https://pan.baidu.com/s/1UhDrC7iOMWiYaQiHk3acdw 提取码:yjgv
## onnx模型转换说明
+
### 一、如何获得模型
+
* 您可以通过使用paddle全家桶,通过其完善的说明,事无巨细的指引,手把手地让您从零开始,训练自己的模型。
* 或者,您也可以直接去paddle官网下载预先训练好的模型。
* 这个预先训练好的模型叫作train模型,server模型体积较大(准确度高速度慢些),mobile模型体积较小(准确度相对低速度快)。
* det和rec有server和mobile两种模型,cls只有mobile模型。
* 预训练模型下载地址:
-1. [mobile det train](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)
-2. [mobile cls train](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar)
-3. [mobile rec train](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar)
-4. [server det train](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)
-5. [server rec train](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar)
+
+|模型名称|模型简介|配置文件|下载地址|
+| --- | --- | --- | --- |
+|mobile cls train|文本方向分类|rec_chinese_lite_train_v2.0.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) |
+|ch_ppocr_mobile_v2.0_det|超轻量模型,多语种文本检测|ch_det_mv3_db_v2.0.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|
+|ch_ppocr_server_v2.0_det|通用模型,多语种文本检测|ch_det_res18_db_v2.0.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)|
+|ch_ppocr_mobile_v2.0_rec|超轻量模型,简体中英文数字|rec_chinese_lite_train_v2.0.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) |
+|ch_ppocr_server_v2.0_rec|通用模型,简体中英文数字|rec_chinese_common_train_v2.0.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar) |
+|en_number_mobile_v2.0_rec|英文数字|rec_en_number_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_train.tar) |
+|french_mobile_v2.0_rec|法文|rec_french_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_train.tar) |
+|german_mobile_v2.0_rec|德文|rec_german_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_train.tar) |
+|korean_mobile_v2.0_rec|韩文|rec_korean_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_train.tar) |
+|japan_mobile_v2.0_rec|日文|rec_japan_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_train.tar) |
+|it_mobile_v2.0_rec|意大利文|rec_it_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/it_mobile_v2.0_rec_train.tar) |
+|xi_mobile_v2.0_rec|西班牙文|rec_xi_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/xi_mobile_v2.0_rec_train.tar) |
+|pu_mobile_v2.0_rec|葡萄牙文|rec_pu_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/pu_mobile_v2.0_rec_train.tar) |
+|ru_mobile_v2.0_rec|俄罗斯文|rec_ru_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ru_mobile_v2.0_rec_train.tar) |
+|ar_mobile_v2.0_rec|阿拉伯文|rec_ar_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ar_mobile_v2.0_rec_train.tar) |
+|hi_mobile_v2.0_rec|印地文|rec_hi_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/hi_mobile_v2.0_rec_train.tar) |
+|chinese_cht_mobile_v2.0_rec|繁体中文|rec_chinese_cht_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/chinese_cht_mobile_v2.0_rec_train.tar) |
+|ug_mobile_v2.0_rec|维吾尔文|rec_ug_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ug_mobile_v2.0_rec_train.tar) |
+|fa_mobile_v2.0_rec|波斯文|rec_fa_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/fa_mobile_v2.0_rec_train.tar) |
+|ur_mobile_v2.0_rec|乌尔都文|rec_ur_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ur_mobile_v2.0_rec_train.tar) |
+|rs_mobile_v2.0_rec|塞尔维亚文(latin)|rec_rs_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/rs_mobile_v2.0_rec_train.tar) |
+|oc_mobile_v2.0_rec|欧西坦文|rec_oc_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/oc_mobile_v2.0_rec_train.tar) |
+|mr_mobile_v2.0_rec|马拉地文|rec_mr_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/mr_mobile_v2.0_rec_train.tar) |
+|ne_mobile_v2.0_rec|尼泊尔文|rec_ne_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ne_mobile_v2.0_rec_train.tar) |
+|rsc_mobile_v2.0_rec|塞尔维亚文(cyrillic)|rec_rsc_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/rsc_mobile_v2.0_rec_train.tar) |
+|bg_mobile_v2.0_rec|保加利亚文|rec_bg_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/bg_mobile_v2.0_rec_train.tar) |
+|uk_mobile_v2.0_rec|乌克兰文|rec_uk_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/uk_mobile_v2.0_rec_train.tar) |
+|be_mobile_v2.0_rec|白俄罗斯文|rec_be_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/be_mobile_v2.0_rec_train.tar) |
+|te_mobile_v2.0_rec|泰卢固文|rec_te_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/te_mobile_v2.0_rec_train.tar) |
+|ka_mobile_v2.0_rec|卡纳达文|rec_ka_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ka_mobile_v2.0_rec_train.tar) |
+|ta_mobile_v2.0_rec|泰米尔文|rec_ta_lite_train.yml|[训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/ta_mobile_v2.0_rec_train.tar) |
+
* 先把这些模型通过训练或者下载的方式准备好备用。
### 二、准备环境和工具
+
* 开始之前,需要先准备好Linux,paddle的pip包不支持macOS,而windows部署起来麻烦一些,以下说明以deepin20为范例。
* 安装python3: ```sudo apt-get install python3```
-* 安装pip3:
+* 安装pip3:
+
```
wget https://bootstrap.pypa.io/get-pip.py
sudo python3 get-pip.py
```
-* pip安装paddlepaddle(2.0的模型需要2.0版): ```sudo pip3 install paddlepaddle==2.0.0rc1 -i https://mirror.baidu.com/pypi/simple```
-* pip安装转换工具附加依赖: ```sudo pip3 install setuptools tqdm opencv-python imgaug pyclipper -i https://mirror.baidu.com/pypi/simple```
+
+* pip安装paddlepaddle(
+ 2.0的模型需要2.0版): ```sudo pip3 install paddlepaddle==2.0.0rc1 -i https://mirror.baidu.com/pypi/simple```
+*
+
+pip安装转换工具附加依赖: ```sudo pip3 install setuptools tqdm opencv-python imgaug pyclipper -i https://mirror.baidu.com/pypi/simple```
+
* 安装lmdb依赖: ```sudo apt-get install python3-yaml libxml2-dev libxslt1-dev python3-dev```
* pip安装lmdb: ```sudo pip3 install lmdb -i https://mirror.baidu.com/pypi/simple```
* pip安装onnx: ```sudo pip3 install onnx```
-* 下载PaddleOCR(用于预训练模型转paddle推理模型):```git clone https://github.com/PaddlePaddle/PaddleOCR.git```
+* 下载PaddleOCR(用于预训练模型转paddle推理模型):```git clone https://github.com/PaddlePaddle/PaddleOCR.git```
* 安装转换工具库Paddle2ONNX(develop分支,用于paddle推理模型转onnx模型):
+
```shell
-#以下为修复mobile rec转换错误的开发版
-#修复代码还未同步进官方仓库之前,先使用此版本
-git clone https://github.com/Channingss/paddle2onnx.git
-cd paddle2onnx
-git checkout fix_adative_pool
+#官方仓库已修复fix_adative_pool
+git clone https://github.com/PaddlePaddle/Paddle2ONNX.git
+cd Paddle2ONNX
+git checkout develop
sudo python3 setup.py install
-#以下是官方仓库
-#git clone https://github.com/PaddlePaddle/Paddle2ONNX.git
-#cd Paddle2ONNX
+#以下为修复mobile rec转换错误的开发版
+#修复代码还未同步进官方仓库之前,先使用此版本
+#git clone https://github.com/Channingss/paddle2onnx.git
+#cd paddle2onnx
+#git checkout fix_adative_pool
#sudo python3 setup.py install
```
### 三、开始转换
+
* 接下来的操作都在PaddleOCR文件夹里进行
-* PaddleOCR切换到2.0分支(因为预训练模型也是v2.0格式的)
+* PaddleOCR切换到release/2.0分支(因为train训练模型也是v2.0格式的)
+
```shell
git clone https://github.com/PaddlePaddle/PaddleOCR.git
cd PaddleOCR
-git checkout release/2.0-rc1-0
+git checkout release/2.0
```
-* 新建文件夹PaddleOCR/models
-* 把预训练模型解压到models目录,严格按照如下目录结构和文件名放置
+
+* 生成多语言识别模型配置文件
+
+```shell
+#安装模块依赖
+sudo pip3 install pyyaml
+cd PaddleOCR/configs/rec/multi_language
+#生成多语言配置文件
+python3 generate_multi_language_configs.py -l it
+python3 generate_multi_language_configs.py -l xi
+python3 generate_multi_language_configs.py -l pu
+python3 generate_multi_language_configs.py -l ru
+python3 generate_multi_language_configs.py -l ar
+python3 generate_multi_language_configs.py -l hi
+python3 generate_multi_language_configs.py -l chinese_cht
+python3 generate_multi_language_configs.py -l ug
+python3 generate_multi_language_configs.py -l fa
+python3 generate_multi_language_configs.py -l ur
+python3 generate_multi_language_configs.py -l rs
+python3 generate_multi_language_configs.py -l oc
+python3 generate_multi_language_configs.py -l mr
+python3 generate_multi_language_configs.py -l ne
+python3 generate_multi_language_configs.py -l rsc
+python3 generate_multi_language_configs.py -l bg
+python3 generate_multi_language_configs.py -l uk
+python3 generate_multi_language_configs.py -l be
+python3 generate_multi_language_configs.py -l te
+python3 generate_multi_language_configs.py -l ka
+python3 generate_multi_language_configs.py -l ta
```
-PaddleOCR/models
- ├── ch_ppocr_mobile_v2.0_cls_train
- │ ├── best_accuracy.pdopt
- │ ├── best_accuracy.pdparams
- │ ├── best_accuracy.states
- │ └── train.log
- ├── ch_ppocr_mobile_v2.0_det_train
- │ ├── best_accuracy.pdopt
- │ ├── best_accuracy.pdparams
- │ ├── best_accuracy.states
- │ └── train.log
- ├── ch_ppocr_mobile_v2.0_rec_train
- │ ├── best_accuracy.pdparams
- │ └── train.log
- ├── ch_ppocr_server_v2.0_det_train
- │ ├── best_accuracy.pdopt
- │ ├── best_accuracy.pdparams
- │ ├── best_accuracy.states
- │ └── train.log
- ├── ch_ppocr_server_v2.0_rec_train
- │ ├── best_accuracy.pdopt
- │ ├── best_accuracy.pdparams
- │ ├── best_accuracy.states
- │ └── train.log
+
+* 新建文件夹PaddleOCR/inference
+* 把预训练模型解压到inference目录,严格按照如下目录结构和文件名放置
+
+```
+PaddleOCR/inference
+├── ar_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── be_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── bg_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ch_ppocr_mobile_v2.0_cls_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ch_ppocr_mobile_v2.0_det_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ch_ppocr_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdparams
+│ └── train.log
+├── ch_ppocr_server_v2.0_det_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ch_ppocr_server_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── chinese_cht_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── en_number_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── fa_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── french_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── german_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── hi_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── it_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── japan_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ka_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── korean_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── mr_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ne_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── oc_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── pu_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── rs_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── rsc_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ru_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ta_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── te_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ug_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── uk_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+├── ur_mobile_v2.0_rec_train
+│ ├── best_accuracy.pdopt
+│ ├── best_accuracy.pdparams
+│ ├── best_accuracy.states
+│ └── train.log
+└── xi_mobile_v2.0_rec_train
+ ├── best_accuracy.pdopt
+ ├── best_accuracy.pdparams
+ ├── best_accuracy.states
+ └── train.log
```
+
* 把本项目tools文件夹里的export_custom.py复制到PaddleOCR/tools里,把program.py覆盖到PaddleOCR/tools里
* 把本项目tools文件夹里的gen-models.sh复制到PaddleOCR根目录
* 开始转换
+
```shell
cd PaddleOCR
chmod a+x gen-models.sh
./gen-models.sh
```
-* 转换成功后,会在PaddleOCR/models文件夹里生成5个onnx模型
+
+* 转换成功后,会在PaddleOCR/inference文件夹里生成31个onnx模型
diff --git a/tools/export_custom.py b/tools/export_custom.py
index c03ad74..6a79867 100755
--- a/tools/export_custom.py
+++ b/tools/export_custom.py
@@ -36,8 +36,6 @@ def main():
config = load_config(FLAGS.config)
merge_config(FLAGS.opt)
logger = get_logger()
- logger.info('input height is {}'.format(FLAGS.height))
- logger.info('input width is {}'.format(FLAGS.width))
# build post process
post_process_class = build_post_process(config['PostProcess'],
@@ -53,7 +51,11 @@ def main():
model.eval()
save_path = '{}/inference'.format(config['Global']['save_inference_dir'])
- infer_shape = [3, int(FLAGS.height), int(FLAGS.width)]
+
+ infer_shape = [3, int(FLAGS.height), int(FLAGS.width)]
+ if config['Architecture']['model_type'] == "rec":
+ infer_shape = [3, 32, -1]
+
model = to_static(
model,
input_spec=[
diff --git a/tools/gen-models.sh b/tools/gen-models.sh
index 900fb3f..f35260e 100755
--- a/tools/gen-models.sh
+++ b/tools/gen-models.sh
@@ -1,32 +1,110 @@
#!/bin/bash
-# ======ch_ppocr_mobile_v2.0_cls======
-python3 tools/export_custom.py --height 48 --width 192 -c configs/cls/cls_mv3.yml -o Global.pretrained_model=./models/ch_ppocr_mobile_v2.0_cls_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./models/ch_ppocr_mobile_v2.0_cls_infer/
+function trainToOnnx() {
+ python3 tools/export_custom.py --height ${1} \
+ --width ${2} \
+ -c ${3} \
+ -o Global.pretrained_model=./inference/${4}_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/${4}_infer/
-paddle2onnx --model_dir models/ch_ppocr_mobile_v2.0_cls_infer/ --model_filename inference.pdmodel --params_filename inference.pdiparams --save_file models/ch_ppocr_mobile_v2.0_cls_infer.onnx --opset_version 11 --enable_onnx_checker True
+ paddle2onnx --model_dir inference/${4}_infer \
+ --model_filename inference.pdmodel \
+ --params_filename inference.pdiparams \
+ --save_file inference/${4}_infer.onnx \
+ --opset_version 11 \
+ --enable_onnx_checker True
+}
-# ======ch_ppocr_mobile_v2.0_det======
-python3 tools/export_custom.py --height -1 --width -1 -c configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml -o Global.pretrained_model=./models/ch_ppocr_mobile_v2.0_det_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./models/ch_ppocr_mobile_v2.0_det_infer/
+# ch_ppocr_mobile_v2.0_cls
+trainToOnnx 48 192 configs/cls/cls_mv3.yml ch_ppocr_mobile_v2.0_cls
-paddle2onnx --model_dir models/ch_ppocr_mobile_v2.0_det_infer/ --model_filename inference.pdmodel --params_filename inference.pdiparams --save_file models/ch_ppocr_mobile_v2.0_det_infer.onnx --opset_version 11 --enable_onnx_checker True
+# ch_ppocr_mobile_v2.0_det
+trainToOnnx -1 -1 configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml ch_ppocr_mobile_v2.0_det
-# ======ch_ppocr_mobile_v2.0_rec======
-python3 tools/export_custom.py --height 32 --width -1 -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model=./models/ch_ppocr_mobile_v2.0_rec_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./models/ch_ppocr_mobile_v2.0_rec_infer/
+# ch_ppocr_server_v2.0_det
+trainToOnnx -1 -1 configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml ch_ppocr_server_v2.0_det
-paddle2onnx --model_dir models/ch_ppocr_mobile_v2.0_rec_infer/ --model_filename inference.pdmodel --params_filename inference.pdiparams --save_file models/ch_ppocr_mobile_v2.0_rec_infer.onnx --opset_version 11 --enable_onnx_checker True
+# ch_ppocr_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml ch_ppocr_mobile_v2.0_rec
+# ch_ppocr_server_v2.0_rec
+trainToOnnx 32 -1 configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml ch_ppocr_server_v2.0_rec
-# ======ch_ppocr_server_v2.0_det======
-python3 tools/export_custom.py --height -1 --width -1 -c configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml -o Global.pretrained_model=./models/ch_ppocr_server_v2.0_det_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./models/ch_ppocr_server_v2.0_det_infer/
+# en_number_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_en_number_lite_train.yml en_number_mobile_v2.0_rec
-paddle2onnx --model_dir models/ch_ppocr_server_v2.0_det_infer/ --model_filename inference.pdmodel --params_filename inference.pdiparams --save_file models/ch_ppocr_server_v2.0_det_infer.onnx --opset_version 11 --enable_onnx_checker True
+# french_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_french_lite_train.yml french_mobile_v2.0_rec
+# german_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_german_lite_train.yml german_mobile_v2.0_rec
-# ======ch_ppocr_server_v2.0_rec======
-python3 tools/export_custom.py --height 32 --width -1 -c configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml -o Global.pretrained_model=./models/ch_ppocr_server_v2.0_rec_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./models/ch_ppocr_server_v2.0_rec_infer/
+# korean_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_korean_lite_train.yml korean_mobile_v2.0_rec
-paddle2onnx --model_dir models/ch_ppocr_server_v2.0_rec_infer/ --model_filename inference.pdmodel --params_filename inference.pdiparams --save_file models/ch_ppocr_server_v2.0_rec_infer.onnx --opset_version 11 --enable_onnx_checker True
+# japan_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_japan_lite_train.yml japan_mobile_v2.0_rec
+# it_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_it_lite_train.yml it_mobile_v2.0_rec
+# xi_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_xi_lite_train.yml xi_mobile_v2.0_rec
+
+# pu_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_pu_lite_train.yml pu_mobile_v2.0_rec
+
+# ru_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ru_lite_train.yml ru_mobile_v2.0_rec
+
+# ar_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ar_lite_train.yml ar_mobile_v2.0_rec
+
+# hi_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_hi_lite_train.yml hi_mobile_v2.0_rec
+
+# chinese_cht_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_chinese_cht_lite_train.yml chinese_cht_mobile_v2.0_rec
+
+# ug_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ug_lite_train.yml ug_mobile_v2.0_rec
+
+# fa_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_fa_lite_train.yml fa_mobile_v2.0_rec
+
+# ur_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ur_lite_train.yml ur_mobile_v2.0_rec
+
+# rs_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_rs_lite_train.yml rs_mobile_v2.0_rec
+
+# oc_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_oc_lite_train.yml oc_mobile_v2.0_rec
+
+# mr_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_mr_lite_train.yml mr_mobile_v2.0_rec
+
+# ne_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ne_lite_train.yml ne_mobile_v2.0_rec
+
+# rsc_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_rsc_lite_train.yml rsc_mobile_v2.0_rec
+
+# bg_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_bg_lite_train.yml bg_mobile_v2.0_rec
+
+# uk_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_uk_lite_train.yml uk_mobile_v2.0_rec
+
+# be_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_be_lite_train.yml be_mobile_v2.0_rec
+
+# te_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_te_lite_train.yml te_mobile_v2.0_rec
+
+# ka_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ka_lite_train.yml ka_mobile_v2.0_rec
+
+# ta_mobile_v2.0_rec
+trainToOnnx 32 -1 configs/rec/multi_language/rec_ta_lite_train.yml ta_mobile_v2.0_rec
diff --git a/tools/program.py b/tools/program.py
index 802365b..b8c94db 100755
--- a/tools/program.py
+++ b/tools/program.py
@@ -133,7 +133,7 @@ def check_gpu(use_gpu):
"model on CPU"
try:
- if use_gpu and not paddle.fluid.is_compiled_with_cuda():
+ if use_gpu and not paddle.is_compiled_with_cuda():
print(err)
sys.exit(1)
except Exception as e:
@@ -181,11 +181,11 @@ def train(config,
if 'start_epoch' in best_model_dict:
start_epoch = best_model_dict['start_epoch']
else:
- start_epoch = 0
+ start_epoch = 1
- for epoch in range(start_epoch, epoch_num):
- if epoch > 0:
- train_dataloader = build_dataloader(config, 'Train', device, logger)
+ for epoch in range(start_epoch, epoch_num + 1):
+ train_dataloader = build_dataloader(
+ config, 'Train', device, logger, seed=epoch)
train_batch_cost = 0.0
train_reader_cost = 0.0
batch_sum = 0
@@ -214,7 +214,7 @@ def train(config,
stats['lr'] = lr
train_stats.update(stats)
- if cal_metric_during_train: # onlt rec and cls need
+ if cal_metric_during_train: # only rec and cls need
batch = [item.numpy() for item in batch]
post_result = post_process_class(preds, batch[1])
eval_class(post_result, batch)
@@ -240,21 +240,21 @@ def train(config,
# eval
if global_step > start_eval_step and \
(global_step - start_eval_step) % eval_batch_step == 0 and dist.get_rank() == 0:
- cur_metirc = eval(model, valid_dataloader, post_process_class,
+ cur_metric = eval(model, valid_dataloader, post_process_class,
eval_class)
- cur_metirc_str = 'cur metirc, {}'.format(', '.join(
- ['{}: {}'.format(k, v) for k, v in cur_metirc.items()]))
- logger.info(cur_metirc_str)
+ cur_metric_str = 'cur metric, {}'.format(', '.join(
+ ['{}: {}'.format(k, v) for k, v in cur_metric.items()]))
+ logger.info(cur_metric_str)
# logger metric
if vdl_writer is not None:
- for k, v in cur_metirc.items():
+ for k, v in cur_metric.items():
if isinstance(v, (float, int)):
vdl_writer.add_scalar('EVAL/{}'.format(k),
- cur_metirc[k], global_step)
- if cur_metirc[main_indicator] >= best_model_dict[
+ cur_metric[k], global_step)
+ if cur_metric[main_indicator] >= best_model_dict[
main_indicator]:
- best_model_dict.update(cur_metirc)
+ best_model_dict.update(cur_metric)
best_model_dict['best_epoch'] = epoch
save_model(
model,
@@ -265,7 +265,7 @@ def train(config,
prefix='best_accuracy',
best_model_dict=best_model_dict,
epoch=epoch)
- best_str = 'best metirc, {}'.format(', '.join([
+ best_str = 'best metric, {}'.format(', '.join([
'{}: {}'.format(k, v) for k, v in best_model_dict.items()
]))
logger.info(best_str)
@@ -296,7 +296,7 @@ def train(config,
prefix='iter_epoch_{}'.format(epoch),
best_model_dict=best_model_dict,
epoch=epoch)
- best_str = 'best metirc, {}'.format(', '.join(
+ best_str = 'best metric, {}'.format(', '.join(
['{}: {}'.format(k, v) for k, v in best_model_dict.items()]))
logger.info(best_str)
if dist.get_rank() == 0 and vdl_writer is not None:
@@ -325,13 +325,13 @@ def eval(model, valid_dataloader, post_process_class, eval_class):
eval_class(post_result, batch)
pbar.update(1)
total_frame += len(images)
- # Get final metirc,eg. acc or hmean
- metirc = eval_class.get_metric()
+ # Get final metric,eg. acc or hmean
+ metric = eval_class.get_metric()
pbar.close()
model.train()
- metirc['fps'] = total_frame / total_time
- return metirc
+ metric['fps'] = total_frame / total_time
+ return metric
def preprocess(is_train=False):