chineseocr / darknet-ocr Goto Github PK

看到更新了ocr识别的模型，马上拿来用一下，自己是windows10的系统，也没有gpu，但是在编译的时候出错了，我电脑上装的MinGW来make的，请问windows系统应该怎么编译呢 是不是要改一下makefile文件，谢谢哈，或者是否可以将build好的library也直接传上来下载后直接用呢
E:\pycharm_workspace\darknet-ocr-master\darknet>make
make: *** No rule to make target 'obj', needed by 'all'. Stop.

@wenlihaoyu @chineseocr 麻烦问一下有身份证、营业执照、护照的相关数据集吗？谢谢

$ python3 app.py 8080
app.py:81: SyntaxWarning: "is not" with a literal. Did you mean "!="?
if 'uuid' is not None and (url is not None or imgString is not None):
Traceback (most recent call last):
File "app.py", line 14, in
from dnn.text import detect_lines
File "/Users/akore/Downloads/darknet-ocr-master/dnn/text.py", line 13, in
from dnn.darknet import load_net,predict_image,array_to_image
File "/Users/akore/Downloads/darknet-ocr-master/dnn/darknet.py", line 112, in
predict_image =lib.network_predict_image_vgg
File "/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/ctypes/init.py", line 382, in getattr
func = self.getitem(name)
File "/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/ctypes/init.py", line 387, in getitem
func = self._FuncPtr((name_or_ordinal, self))
AttributeError: dlsym(0x7ff287681f10, network_predict_image_vgg): symbol not found

新手小白，求问 怎么编译CPU版本？多谢

大佬你的数据集是生成么，如果是请指点一下生成所需要考虑到的细节因素，还有您使用字符集是多少个的、有没有使用专业语料库之类的

223.73.120.86:30801 - - [14/Oct/2019 11:43:45] "HTTP/1.1 POST /text" - 500 Internal Server Error
Traceback (most recent call last):
  File "/opt/miniconda3/envs/darknet-ocr/lib/python3.6/site-packages/web/application.py", line 257, in process
    return self.handle()
  File "/opt/miniconda3/envs/darknet-ocr/lib/python3.6/site-packages/web/application.py", line 248, in handle
    return self._delegate(fn, self.fvars, args)
  File "/opt/miniconda3/envs/darknet-ocr/lib/python3.6/site-packages/web/application.py", line 488, in _delegate
    return handle_class(cls)
  File "/opt/miniconda3/envs/darknet-ocr/lib/python3.6/site-packages/web/application.py", line 466, in handle_class
    return tocall(*args)
  File "app.py", line 83, in POST
    res = job(uuid,url,imgString,iscut,isclass,billModel,ip)
  File "app.py", line 34, in job
    boxes,scores = detect_lines(image,scale=scale,maxScale=maxScale)
  File "/opt/darknet-ocr/dnn/text.py", line 48, in detect_lines
    scores,boxes,rate,w,h = detect_box(image,scale,maxScale)
  File "/opt/darknet-ocr/dnn/text.py", line 22, in detect_box
    out  = textNet.forward(outputName)[0]
cv2.error: OpenCV(4.0.0) /io/opencv/modules/dnn/src/layers/concat_layer.cpp:95: error: (-201:Incorrect size of input array) Inconsistent shape for ConcatLayer in function 'getMemoryShapes'

环境:
ubuntu 16.04 amd64

<1>在src中的network.cu中1005行：
void sync_n
ets(network **nets, int n, int interval)
正确应该为：void sync_nets(network **nets, int n, int interval)
<2>在src中的maxpool_layer_kernels.cu中36,37行：
int cur_h = h_offset + istride[0] + l;
int cur_w = w_offset + jstride[1] + m;
正确应为：
int cur_h = h_offset + istrideh + l;
int cur_w = w_offset + jstridew + m;
<3>在src中的maxpool_layer_kernels.cu中55行：
int area = (size-1)/stride;
这个stride到底是strideh 还是stridew ，我也蒙圈了

cd darknet && cp Makefile-GPU Makefile && make
make: *** No rule to make target 'obj', needed by 'all'. Stop.

至于分类模型， dnn'跑报错：Relu不支持，作者确定测试过opencv dnn模块可以运行ocr吗？

127.0.0.1:48050 - - [14/Jan/2020 18:12:47] "HTTP/1.1 POST /text" - 500 Internal Server Error
Traceback (most recent call last):
File "/home/ubuntu/.local/lib/python3.5/site-packages/web/application.py", line 289, in process
return self.handle()
File "/home/ubuntu/.local/lib/python3.5/site-packages/web/application.py", line 280, in handle
return self._delegate(fn, self.fvars, args)
File "/home/ubuntu/.local/lib/python3.5/site-packages/web/application.py", line 530, in _delegate
return handle_class(cls)
File "/home/ubuntu/.local/lib/python3.5/site-packages/web/application.py", line 508, in handle_class
return tocall(*args)
File "app.py", line 66, in POST
data = json.loads(data)
File "/usr/lib/python3.5/json/init.py", line 312, in loads
s.class.name))
TypeError: the JSON object must be str, not 'bytes'

127.0.0.1:48236 - - [14/Jan/2020 18:13:35] "HTTP/1.1 POST /text" - 500 Internal Server Error

Traceback (most recent call last):
File "app.py", line 14, in
from dnn.text import detect_lines
File "/home/wanghe/darknet-ocr/dnn/text.py", line 13, in
from dnn.darknet import load_net,predict_image,array_to_image
File "/home/wanghe/darknet-ocr/dnn/darknet.py", line 112, in
predict_image =lib.network_predict_image_vgg
File "/usr/lib64/python3.6/ctypes/init.py", line 356, in getattr
func = self.getitem(name)
File "/usr/lib64/python3.6/ctypes/init.py", line 361, in getitem
func = self._FuncPtr((name_or_ordinal, self))
AttributeError: ../darknet/libdarknet.so: undefined symbol: network_predict_image_vgg

this is my error,i git clone darknet and make it,then exec /bin/python3 app.py 8080
i make the darknet-ocr,but it does not work,prompt make: *** There are no rules to create the target "obj" required by "all". stop

您好！请问如何提高检测速度呢？

win10下CPU模式，python3.6.5，已编译cpu版makefile

请问CPU模式下，config.py设置GPU为False，是否需要修改ocr.py第19行 if 1: 为 if gpu: ？？？

若不修改，调用predict_darknet 和 predict_cpu均无任何输出，无报错信息，程序就直接结束

若修改，输出结果均为乱码，如下图。

您好！您用了多少数据进行的训练，有训练数据吗？

I run app.py but it is a error, how to solve it? Thanks

OSError: dlopen(darknet/libdarknet.so, 10): image not found

1、采用cygwin make好darknet并部署
2、pip安装了库
python app.py 8080 后，命令行直接退出，无任何提示，浏览器无法访问，
请问有人遇到过吗？

测试了几个系统，识别结果都出现乱码。
请问下开发时候的系统环境。

推荐Apache 或 MIT协议

以下是结果：
序号 值
0 海政面憶九拳衝釘釧貝ャ割握視晩キ稼熔材井釘窒×芭赤稼村政朗台憶造_鎖劣條壊胆療塚政あ海●五困

请问日语模型是不是还不能使用

感觉启动后好多问题都是在templates/text.htm 中的python格式各种语法报错
我的是mac的3.8的python，跑起来比较艰难

您好！
请问您的训练代码打算什么时候开源 谢谢

如题

Ubuntu 16.04
python3.7
无GPU，编译好了darknet，config也设置了darkRoot
测试运行dnn/ocr.py 报错Unknown error 47391493
定位错误是在dnn/ocr.py 23行 ocrNet = load_net(ocrPath.replace('.weights','.cfg').encode(),ocrPath.encode(), 0)

大神，您好：

大概测试了一下，cpu版本是没有问题的，
但gpu版本在运行时，会报：: Unknown error 1362440885 的错误，程序退出。
初步定位了下，是在：dnn/text.py 的14行：
textNet = load_net(textPath.replace('.weights','.cfg').encode(),textPath.encode(), 0)
运行这一步时，报的错， 没有其他输出信息。
请问这个是什么原因导致的? 谢谢。

make: *** No rule to make target 'obj', needed by 'all'. Stop.

array([[ 69.375 , 70.10526316, 244.63815789, 68.64473684,
244.63815789, 101.50657895, 69.375 , 102.23684211],
[ 69.375 , 163.57894737, 361.48026316, 164.30921053,
361.48026316, 198.63157895, 69.375 , 197.90131579],
[ 57.69078947, 341.03289474, 595.16447368, 342.49342105,
595.16447368, 378.27631579, 57.69078947, 376.81578947],
[361.48026316, 249.75 , 384.84868421, 249.75 ,
384.84868421, 274.57894737, 361.48026316, 274.57894737],
[ 57.69078947, 213.96710526, 420.63157895, 221.26973684,
419.90131579, 254.86184211, 56.96052632, 247.55921053],
[ 45.27631579, 119.76315789, 314.74342105, 116.11184211,
315.47368421, 148.97368421, 46.00657895, 152.625 ],
[116.11184211, 248.28947368, 303.78947368, 252.67105263,
303.05921053, 284.07236842, 115.38157895, 280.42105263]])

这个坐标顺序是什么呢？

请问门店广告字 和数字标签能识别吗？

数字 1~10， 占满图片可以识别吗？

你好，请问英文OCR模型能否上传测试？谢谢

IE， Chrome，Firefox 都无法正常显示，也无法识别图片，自带的一些图片也无法识别。
上传图片，点击识别，提示异常

请及时看看问题

--2020-01-08 23:18:08-- http://www.chineseocr.com:9990/static/models/darknet-ocr/models/ocr/chinese/ocr.weights
Resolving www.chineseocr.com... 59.110.234.163
Connecting to www.chineseocr.com|59.110.234.163|:9990... connected.
HTTP request sent, awaiting response... 404 File not found
2020-01-08 23:18:08 ERROR 404: File not found.

请问有没有纯数字的ocr模型，主要用于识别印刷体照片扫描件。
或则能否提供生成数据的方法

AttributeError: 'cv2.dnn_Net' object has no attribute 'getUnconnectedOutLayersNames'

版主您好，我點進模型文建下載時百度網盤說網頁不存在了，不知道是不是地區的關係從來沒有一次從百度網盤下在成功，方便提供其他的載點嗎，謝謝!

env:
python 3.7.1
web.py==0.40.dev0

problem:
return Template(open(path).read(), filename=path, **self._keywords)
UnicodeDecodeError: 'gbk' codec can't decode byte 0xaf in position 410: illegal multibyte sequence

temporary solution

open the file lib\site-packages\web\template.py

old: return Template(open(path).read(), filename=path, **self._keywords)
new: return Template(open(path,encoding='utf-8').read(), filename=path, **self._keywords)

作者你好，想问一下ocr.cfg该如何使用呢，训练数据是怎样标注的？

素材：

(base) xxxx@ubuntu:~/darknet-ocr-master$ python app.py 8080
Traceback (most recent call last):
File "app.py", line 14, in
from dnn.text import detect_lines
File "/home/xxx/darknet-ocr-master/dnn/text.py", line 13, in
from dnn.darknet import load_net,predict_image,array_to_image
File "/home/xxx/darknet-ocr-master/dnn/darknet.py", line 112, in
predict_image =lib.network_predict_image_vgg
File "/home/xxx/anaconda3/lib/python3.6/ctypes/init.py", line 357, in getattr
func = self.getitem(name)
File "/home/xxx/anaconda3/lib/python3.6/ctypes/init.py", line 362, in getitem
func = self._FuncPtr((name_or_ordinal, self))
AttributeError: ../darknet/libdarknet.so: undefined symbol: network_predict_image_vgg

多方查找资料，也多次编译darknet但问题依然，请指导，非常感谢

python web 8080 段端口在哪个文件修改？

怎样做数据标注，和训练

请问作者在这个darknet-ocr项目中是准备放弃yolo3来检测文字了吗？
如果确实如此，请问是出于什么考虑，方便介绍一下吗，感谢。

访问识别页面，上传图片之后，出来的结果是乱码，序号和值是正常的，leable里面的数据是乱码

Windows10, Python 3.5.2 直接上传本地图片 /test/orc.jpeg
提示：识别异常
cmd窗口 log 如下：

python app.py 8080
http://0.0.0.0:8080/
127.0.0.1:58197 - - [13/Oct/2019 21:18:50] "HTTP/1.1 GET /text" - 200 OK
127.0.0.1:58200 - - [13/Oct/2019 21:18:50] "HTTP/1.1 GET /static/css/main.css" - 304 Not Modified
127.0.0.1:58198 - - [13/Oct/2019 21:18:50] "HTTP/1.1 GET /static/js/jquery.js" - 304 Not Modified
127.0.0.1:58199 - - [13/Oct/2019 21:18:50] "HTTP/1.1 GET /static/js/helps.js" - 304 Not Modified
127.0.0.1:58197 - - [13/Oct/2019 21:18:50] "HTTP/1.1 GET /static/img/loading.gif" - 304 Not Modified
Traceback (most recent call last):
File "C:\Users\admin\AppData\Local\Programs\Python\Python35\lib\site-packages\web.py-0.40-py3.5.egg\web\application.py", line 289, in process
return self.handle()
File "C:\Users\admin\AppData\Local\Programs\Python\Python35\lib\site-packages\web.py-0.40-py3.5.egg\web\application.py", line 280, in handle
return self._delegate(fn, self.fvars, args)
File "C:\Users\admin\AppData\Local\Programs\Python\Python35\lib\site-packages\web.py-0.40-py3.5.egg\web\application.py", line 530, in delegate
return handle_class(cls)
File "C:\Users\admin\AppData\Local\Programs\Python\Python35\lib\site-packages\web.py-0.40-py3.5.egg\web\application.py", line 508, in handle_class
return tocall(*args)
File "app.py", line 73, in POST
data = json.loads(data)
File "C:\Users\admin\AppData\Local\Programs\Python\Python35\lib\json_init.py", line 312, in loads
s.class.name))
TypeError: the JSON object must be str, not 'bytes'

127.0.0.1:58197 - - [13/Oct/2019 21:19:00] "HTTP/1.1 POST /text" - 500 Internal Server Error

采用cpu 识别结果正常，采用gpu则没有识别到。无异常log输出。环境为多卡gtx1080ti GPU，ubuntu16

请问，您训练师印刷体文字和手写体文字的比例大概是多少，还有您数据集里有自己生成数据的比例大概有多少，谢谢～～

Python version 3.7.1, web.py==0.40.dev0
problem
I got this problem above,

temporary solution
open this file lib/site-packages/web/httpserver.py

Add:
self.directory = os.getcwd()
After:
self.start_response = start_response

访问这个地址返回 404
http://www.chineseocr.com:9990/static/models/darknet-ocr/models/ocr/chinese/ocr.weights

如标题，文字区域捕捉准确率与识别效果方面，感觉chineseocr项目更好。