optimize PaddleOCR #3

LDOUBLEV · 2020-05-11T07:37:08Z

add checkpoints in yml
add doc directory
supporting sort predict box
add detection fluid inference
masked memory optimization and ir optimization options in utility.py

dyning · 2020-05-11T08:09:48Z

doc/detection.md

+
+## 3.2 快速启动训练
+
+首先下载pretrain model，目前支持两种backbone，分别是MobileNetV3、ResNet50，您可以根据需求使用PaddleClas中的模型更换


需要再详细说明下网络结构，MobileNetV3有10个，ResNet50是ResNet50_vd

dyning · 2020-05-11T08:10:34Z

doc/detection.md

+
+PaddleOCR计算三个OCR检测相关的指标，分别是：Precision、Recall、Hmean。
+
+运行如下代码，根据配置文件det_db_mv3.yml中save_res_path指定的测试集检测结果文件，计算评估指标。


代码里面需要注明评估来源于DB repo

dyning · 2020-05-11T08:11:08Z

doc/detection.md

+运行如下代码，根据配置文件det_db_mv3.yml中save_res_path指定的测试集检测结果文件，计算评估指标。
+
+```
+python3 tools/eval.py -c configs/det/det_db_mv3.yml  -o checkpoints ./output/best_accuracy


这块为什么使用checkpoints，而不是pretrainedweights？

pretrained weights是加载的backbone的模型参数，在评估或者测试时要加载训练好的参数文件，换成checkpoints是不是更能区分些？

dyning · 2020-05-11T08:11:54Z

doc/installation.md

+
+我们提供了PaddleOCR开发环境的docker，您可以pull我们提供的docker运行PaddleOCR的环境。
+
+1. 准备docker环境。第一次使用这个镜像，会自动下载该镜像，请耐心等待。


自动下载删掉，需要让用户操作下载吧

换了一个官方镜像，测试了可以PaddleOCR可以正常运行，不需要用户再去手动下载。

dyning · 2020-05-11T08:13:50Z

tools/infer/utility.py

@@ -99,14 +99,15 @@ def create_predictor(args, mode):
        config.disable_gpu()

    config.disable_glog_info()
-    config.switch_ir_optim(args.ir_optim)
+    # config.switch_ir_optim(args.ir_optim)


直接删了吧，包括trt和fp16的参数

dyning · 2020-05-12T09:04:13Z

README.md

-OCR algorithms with PaddlePaddle （still under develop)
+
+# 简介
+PaddleOCR旨在打造一套丰富、领先、且实用的文字检测、识别模型/工具库，助力使用者训练出更好的模型，并应用落地。


OCR工具库

dyning · 2020-05-12T09:05:21Z

README.md

+
+## 文档教程
+- [快速安装](./doc/installation.md)
+- [文本识别模型训练/评估/预测](./doc/detection.md)


修改目录

添加快速开始
export model

dyning · 2020-05-12T09:10:58Z

README.md

+- [文本识别模型训练/评估/预测](./doc/detection.md)
+- [文本预测模型训练/评估/预测](./doc/recognition.md)
+
+## 特性：


特性上移

dyning · 2020-05-12T09:12:00Z

README.md

+- [SAST](https://arxiv.org/abs/1908.05498)
+
+算法效果：
+|模型|骨干网络|数据集|Hmean|


表格里添加模型引用，然后todo

数据集去掉

dyning · 2020-05-12T09:15:51Z

configs/det/det_db_mv3.yml

@@ -12,6 +12,7 @@ Global:
  image_shape: [3, 640, 640]
  reader_yml: ./configs/det/det_db_icdar15_reader.yml
  pretrain_weights: ./pretrain_models/MobileNetV3_pretrained/MobileNetV3_large_x0_5_pretrained/
+  checkpoints:


确认对不对？

dyning · 2020-05-12T09:16:36Z

doc/detection.md

@@ -0,0 +1,78 @@
+# 文字检测
+
+本节以icdar15数据集为例，介绍PaddleOCR中检测模型的使用方式。


使用方式展开

dyning · 2020-05-12T09:18:14Z

doc/detection.md

+" 图像文件名                    json.dumps编码的图像标注信息"
+ch4_test_images/img_61.jpg    [{"transcription": "MASA", "points": [[310, 104], [416, 141], [418, 216], [312, 179]], ...}]
+```
+json.dumps编码前的图像标注信息是包含多个字典的list，字典中的points表示文本框的位置，如果您想在其他数据集上训练PaddleOCR,


说明检测不需要文本信息，points 四点

x1,y1,x2,y2

dyning · 2020-05-12T09:21:12Z

doc/installation.md

+pip3 install -r requirements.txt
+```
+
+## 快速运行


快速运行删掉

dyning · 2020-05-12T09:24:34Z

tools/infer_det.py

+# NOTE(paddle-dev): All of these flags should be
+# set before `import paddle`. Otherwise, it would
+# not take any effect.
+set_paddle_flags(


暂定不删，等更新1.8代码时更新

dyning · 2020-05-12T09:25:05Z

tools/infer_det.py

+
+
+def simple_reader(img_file, config):
+    imgs_lists = []


读图使用utility的独立函数

Done，读图函数封装了一个函数，位置在tools/infer/utility.py-get_image_file_list()

update2020-7-20

test=develop

* update code & doc

add relative path

LDOUBLEV added 2 commits May 11, 2020 15:27

add doc、infer_det.py、requirments.txt

3f2d384

fix tools/infer/utility.py

af39902

dyning reviewed May 11, 2020

View reviewed changes

LDOUBLEV added 2 commits May 11, 2020 19:59

update detection doc and infer code

268ed03

update doc and requirments

f7bea71

LDOUBLEV changed the title ~~optimizer PaddleOCR~~ optimize PaddleOCR May 12, 2020

LDOUBLEV added 2 commits May 12, 2020 15:04

update readme and doc

561c544

update readme

e91f370

dyning reviewed May 12, 2020

View reviewed changes

LDOUBLEV added 3 commits May 12, 2020 19:40

fix problems refer comments

5b4675e

fix problems responding to inference

b2e2bb9

remove README for now

ab8fd7d

dyning approved these changes May 12, 2020

View reviewed changes

dyning merged commit ed4b270 into PaddlePaddle:develop May 12, 2020

dyning pushed a commit that referenced this pull request Jul 24, 2020

Merge pull request #3 from PaddlePaddle/develop

e0fa21b

update2020-7-20

zhouyongxyz mentioned this pull request Sep 16, 2020

MKLDNN 预测多线程创建多实例出现Segmentation fault #731

Closed

adigest mentioned this pull request Oct 15, 2020

Fatal signal 11 (SIGSEGV), code 1 (SEGV_MAPERR) #944

Closed

rp-koayst mentioned this pull request Jan 26, 2021

How to enable GPU ? #1814

Closed

Yaoxingtian mentioned this pull request Apr 13, 2021

AssertionError: Variable Shape not match, #2479

Closed

idreamerhx mentioned this pull request May 28, 2021

PaddleLite deploy lite core dump #2962

Closed

BillDior pushed a commit to BillDior/PaddleOCR that referenced this pull request Aug 13, 2021

resize images in README.md (PaddlePaddle#3)

5f75fd4

test=develop

bnu-bily mentioned this pull request Jan 18, 2022

内网的情况下怎么执行这一步 #5274

Closed

linkewei0580 mentioned this pull request Jan 26, 2022

PaddleOCR 2.4 检测某个图片的时候core了 #5359

Closed

linkewei0580 mentioned this pull request Mar 4, 2022

paddleocr2.4 core #5632

Closed

qiuming-93 mentioned this pull request Jun 7, 2022

多线程调用C++推理库进行OCR推理时出现程序崩溃 #6512

Closed

an1018 pushed a commit to an1018/PaddleOCR that referenced this pull request Aug 17, 2022

Update api (PaddlePaddle#3)

3b3a509

* update code & doc

This was referenced Jan 11, 2023

多线程调用C++推理库进行OCR推理时程序会崩溃！！！！ #8823

Closed

多线程调用C++推理库进行RNN算子崩溃问题！！！！ PaddlePaddle/Paddle#49737

Open

ChenNima added a commit to ChenNima/PaddleOCR that referenced this pull request May 19, 2023

upgrade-paddle (PaddlePaddle#3)

0690424

old-steel mentioned this pull request Jun 5, 2023

DBnet模型在NPU上跑八卡卡死 #10095

Closed

mxihan pushed a commit to mxihan/PaddleOCR that referenced this pull request Jul 4, 2023

Merge pull request PaddlePaddle#3 from lewangdev/fastapi

270aac3

add relative path

xuxiansheng2018 mentioned this pull request Jul 18, 2023

训练det报WARNING #10427

Closed

cty-ai mentioned this pull request Nov 22, 2023

cpp版本识别重建表格时崩溃 #11290

Open

richardgohth mentioned this pull request Mar 19, 2024

Where to download libpaddle_inference.so version 2.7 for ubuntu 20? #11763

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

optimize PaddleOCR #3

optimize PaddleOCR #3

LDOUBLEV commented May 11, 2020 •

edited

Loading

dyning May 11, 2020

dyning May 11, 2020

LDOUBLEV May 11, 2020

dyning May 11, 2020

LDOUBLEV May 11, 2020

dyning May 11, 2020

LDOUBLEV May 11, 2020

dyning May 11, 2020

LDOUBLEV May 11, 2020

dyning May 12, 2020

dyning May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020

dyning May 12, 2020

LDOUBLEV May 12, 2020 •

edited

Loading


		## 3.2 快速启动训练

		首先下载pretrain model，目前支持两种backbone，分别是MobileNetV3、ResNet50，您可以根据需求使用PaddleClas中的模型更换


		PaddleOCR计算三个OCR检测相关的指标，分别是：Precision、Recall、Hmean。

		运行如下代码，根据配置文件det_db_mv3.yml中save_res_path指定的测试集检测结果文件，计算评估指标。


		我们提供了PaddleOCR开发环境的docker，您可以pull我们提供的docker运行PaddleOCR的环境。

		1. 准备docker环境。第一次使用这个镜像，会自动下载该镜像，请耐心等待。

		@@ -0,0 +1,78 @@
		# 文字检测

		本节以icdar15数据集为例，介绍PaddleOCR中检测模型的使用方式。

optimize PaddleOCR #3

optimize PaddleOCR #3

Conversation

LDOUBLEV commented May 11, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

LDOUBLEV May 12, 2020 • edited Loading

Choose a reason for hiding this comment

LDOUBLEV commented May 11, 2020 •

edited

Loading

LDOUBLEV May 12, 2020 •

edited

Loading