senlinuc/caffe_ocr issues and pull requests

#145 - bn层 lr_mult：0

Issue - State: open - Opened by yu1206 4 months ago

#144 - 求更新网盘链接

Issue - State: open - Opened by chungexintudi almost 2 years ago - 2 comments

#143 - 百度链接没一个能用的

Issue - State: open - Opened by vector2myth about 2 years ago

#142 - 中文数据集链接失效，能否更新下？

Issue - State: open - Opened by ViviKing414 over 2 years ago

#141 - SynthText Data

Issue - State: open - Opened by laofeiwei almost 3 years ago

#140 - 编译报错

Issue - State: open - Opened by ctgushiwei over 3 years ago

#139 - 合作交流

Issue - State: open - Opened by mymutouren almost 4 years ago

#138 - dbtrain

Issue - State: open - Opened by zzcqinag about 4 years ago

#137 - c++调用caffemodel问题

Issue - State: open - Opened by xiaotaia about 4 years ago

#136 - 训练数据集图片数量比train.txt里面的标签数量多了1个。

Issue - State: open - Opened by jjz-learning about 4 years ago - 3 comments

#135 - opencv加载caffe

Issue - State: open - Opened by amandazw over 4 years ago

#134 - 数据问题一直不行，求解答

Issue - State: open - Opened by xiaotaia over 4 years ago

#133 - ctcloss = 0 (* 1 = 0 loss)

Issue - State: closed - Opened by Ryansanity over 4 years ago

#132 - 如何添加未定义的新层？

Issue - State: open - Opened by EdwardVincentMa over 4 years ago

#131 - 大佬好！关于 train.prototxt 的 label

Issue - State: closed - Opened by EdwardVincentMa over 4 years ago - 5 comments

#130 - 英文词典相关问题

Issue - State: open - Opened by cactuszxr over 4 years ago

#129 - 丢字符的问题如何修改网络？

Issue - State: open - Opened by ManiroXi over 4 years ago

#128 - 请问编译需要使用哪个版本的protobuf？感谢回答

Issue - State: open - Opened by vincezengqiang over 4 years ago

#127 - 请问有人试过用efficientnet网络来做文字识别的么？

Issue - State: open - Opened by pingzi5233 almost 5 years ago

#126 - 含densenet的模型可以转换成tensorflow版吗？

Issue - State: open - Opened by doublegssc almost 5 years ago

#125 - 合成的中文数据百度盘已的失效

Issue - State: open - Opened by waji3000 almost 5 years ago - 2 comments

#124 - 准确率是单行全对的准确率?

Issue - State: open - Opened by lightning0016 almost 5 years ago

#123 - 为何要更换为junhyukoh实现的lstm版本

Issue - State: open - Opened by zhangsong8 about 5 years ago

#122 - 为什么测试结果表明resnet18比resnet50效果好

Issue - State: open - Opened by mengpengfei about 5 years ago

#121 - 为什么用双层blstm，单层blstm不会收敛？

Issue - State: closed - Opened by mengpengfei about 5 years ago - 2 comments

#120 - loss不收敛

Issue - State: open - Opened by AnMoran about 5 years ago

#119 - 中文字典在哪里

Issue - State: open - Opened by huitang about 5 years ago - 5 comments

#118 - 您好，我配置标签时这样的

Issue - State: closed - Opened by AnMoran about 5 years ago - 1 comment

#117 - 从词典随机生成数据训练不收敛

Issue - State: closed - Opened by wenlive about 5 years ago - 2 comments

#116 - data_layer.cpp has no member named ‘label_size’

Issue - State: open - Opened by zzcqinag about 5 years ago

#115 - 请问用vs2015打开工程加载caffe失败，是哪里出错了吗

Issue - State: open - Opened by duducode over 5 years ago - 2 comments

#114 - 我创建了普通的densenet 网络，怎么训练效果并不好，你帮我看看，大仙

Issue - State: closed - Opened by sunjunlishi over 5 years ago - 5 comments

#113 - 使用cudnn10.1版本出现问题

Issue - State: open - Opened by zhengqicl over 5 years ago

#112 - ocr_test 选择CPU_ONLY时，找不到libClassificationCPU-MKL.lib

Issue - State: open - Opened by rpf1019 over 5 years ago - 2 comments

#111 - densenet-no-blstm-vertical-feature

Issue - State: open - Opened by stephonjou over 5 years ago - 2 comments

#110 - DenseBlockLayer的 openblas实现

Issue - State: closed - Opened by sunjunlishi over 5 years ago - 15 comments

#109 - 这么多bug的项目竟然还发在这里到处吹

Issue - State: open - Opened by mengpengfei over 5 years ago - 17 comments

#108 - 请教作者，关于opencv调用densenet模型

Issue - State: open - Opened by teresasun521 over 5 years ago - 5 comments

#107 - 字符长度变化较大，在10-50长度，这个数据怎么训练

Issue - State: open - Opened by shaoshuaiche over 5 years ago - 5 comments

#106 - 你好作者，请问均值你是怎么算出来的，

Issue - State: open - Opened by yangninghua over 5 years ago - 1 comment

#105 - 这个怎么那么多bug啊

Issue - State: closed - Opened by mengpengfei over 5 years ago

#104 - 选择哪块显卡运行ocr_test

Issue - State: open - Opened by HouBiaoLiu over 5 years ago

#103 - 能否支持多卡训练？

Issue - State: closed - Opened by Cherry2410 almost 6 years ago

#102 - loss 一直在50，降不下去了，是数据少吗，有30000条数据。

Issue - State: open - Opened by shaoshuaiche almost 6 years ago - 4 comments

#101 - 作者究竟怎么测的，我测试的准确度和速度都比作者表现好得多

Issue - State: closed - Opened by mengpengfei almost 6 years ago

#100 - 支不支持cudnn7，cudnn5比7慢了将近两倍

Issue - State: closed - Opened by mengpengfei almost 6 years ago

#99 - 无法屏蔽glog初始化输出的问题

Issue - State: open - Opened by ZipengFeng almost 6 years ago

#98 - 训练模型时出现status == CTC_STATUS_SUCCESS (3 vs. 0) CTC Error

Issue - State: open - Opened by Echo-byte almost 6 years ago - 3 comments

#97 - 英文数据集训练之后能识别验证码吗？

Issue - State: closed - Opened by mengpengfei almost 6 years ago

#96 - 请教作者senlinuc，为什么使用opencv的readNetFromCaffe调用ocr模型会报错呢？是版本不匹配吗？

Issue - State: open - Opened by teresasun almost 6 years ago - 2 comments

#95 - can you supply a pre-trained model

Issue - State: open - Opened by azuryl almost 6 years ago - 1 comment

#94 - 请问仅CPU如何编译呢？好像没有提供CommonSetting？

Issue - State: closed - Opened by ZipengFeng about 6 years ago

#93 - 解决方案打开的问题

Issue - State: open - Opened by gitmessage about 6 years ago

#92 - 将fc1x层的节点数由5990变为36（因为我只识别数字和大写英文字母），无法训练，loss=nan

Issue - State: open - Opened by HouBiaoLiu about 6 years ago - 4 comments

#91 - densenet-sum-blstm-full-res-blstm_train-val.prototxt只有看到一个标签，怎么样才能对多个标签进行训练

Issue - State: open - Opened by shenjackson about 6 years ago - 1 comment

#90 - 大神，你识别中文是Lexicon-based transcription还是Lexicon-free transcription

Issue - State: open - Opened by bxhandhxb about 6 years ago

#89 - 作者方便把训练的英文字典和数据也分享下吗？

Issue - State: open - Opened by magneter about 6 years ago - 1 comment

#88 - Opencv版本

Issue - State: open - Opened by ffyAI about 6 years ago - 2 comments

#87 - mean_value.txt

Issue - State: open - Opened by panlanlan over 6 years ago

#86 - 中文括号,导致链接错误了

Pull Request - State: open - Opened by wangershi over 6 years ago

#85 - 显示乱码

Issue - State: open - Opened by xqpinitial over 6 years ago

#84 - 数据库生成很好

Issue - State: open - Opened by Caiyou1001 over 6 years ago - 4 comments

#83 - 移植到Linux

Issue - State: closed - Opened by FLHonker over 6 years ago - 1 comment

#82 - 请问英文句子中空格怎样识别？

Issue - State: open - Opened by Jeremyliu7 over 6 years ago - 3 comments

#81 - The meaning of score

Issue - State: open - Opened by nerddd over 6 years ago

#80 - 请教大神如何移植到Android上

Issue - State: open - Opened by xiaomochen520 over 6 years ago - 2 comments

#79 - denseBlock Layer中的initChannel 代表什么含义呢？

Issue - State: closed - Opened by ghost over 6 years ago

#78 - 多值标签制作lmdb问题

Issue - State: open - Opened by suntim over 6 years ago - 3 comments

#77 - ocr_test 有用到均值文件么，未什么我训练的时候准确率很高使用ocr_test测试的时候准确率很低

Issue - State: closed - Opened by ghost over 6 years ago - 2 comments

#76 - Check failed: axis_index < num_axes() (2 vs. 2) axis 2 out of range for 2-D Blob with shape 35 256 (8960)

Issue - State: open - Opened by suntim over 6 years ago - 3 comments

#75 - 验证机的准确率很高，测试集的准确率很低是怎么回事？

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#74 - 使用ImageData层和LMDB方式读取数据有区别吗？

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#73 - 使用orc_test测试的时候，输出label很多，且大部分是0是怎么回事

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#72 - 训练疑问，迭代两万次之后loss卡在30不降低了

Issue - State: closed - Opened by Levi-zhan over 6 years ago - 7 comments

#71 - DenseBlock 出错，cudnnTensorDescriptor_t’ does not name a type

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#70 - warpctcloss层疑问

Issue - State: closed - Opened by prfans over 6 years ago

#69 - 训练疑问

Issue - State: open - Opened by sannyjo over 6 years ago - 1 comment

#68 - 在中文数据集上crnn也是最高,能到99.26,你测试的数据集是30多万哪个么

Issue - State: open - Opened by liny23 over 6 years ago - 18 comments

#67 - 是否ctc对连续数字的识别率不够高？

Issue - State: closed - Opened by ares89 over 6 years ago - 2 comments

#64 - compute_output_shape compute wrong output_shape value

Issue - State: open - Opened by mbtolou over 6 years ago - 1 comment

#60 - 请问下transpose层和permute层有什么区别吗

Issue - State: open - Opened by shaoming20798 almost 7 years ago - 1 comment

#59 - warp_ctc_loss_layer.cpp:56] Check failed: N_ == label_seq* Check failure stack trace: * ->num() (128 vs. 64)

Issue - State: closed - Opened by prfans almost 7 years ago - 29 comments

#56 - DenseBlock_layer.cu(283): error: too many arguments in function call

Issue - State: open - Opened by billyzju almost 7 years ago - 4 comments

#55 - Test Net output 输出的有两个acc,分别指的是什么？

Issue - State: open - Opened by AllenDuane almost 7 years ago - 1 comment

#52 - 关于训练自己的样本，结果有问题

Issue - State: open - Opened by xap1992 almost 7 years ago - 3 comments

#45 - 楼主，为什么我识别后，总是只能识别第一个字，后面就识别不了了。

Issue - State: closed - Opened by GarryWan almost 7 years ago - 2 comments

#42 - 英文模型训练

Issue - State: open - Opened by rabbithui almost 7 years ago - 8 comments

#41 - 样本生成工具能否公开

Issue - State: open - Opened by blacklee5 almost 7 years ago - 10 comments

#36 - 识别英文的模型在哪？

Issue - State: open - Opened by rkshuai about 7 years ago - 8 comments

#33 - 可否公开一下生成中文训练数据的代码？

Issue - State: open - Opened by Ruochen0715 about 7 years ago - 2 comments

#32 - 关于字符长度的问题

Issue - State: open - Opened by AllenDuane about 7 years ago - 2 comments

#31 - ocr_test densenet-no-blstm cpu模式前向识别报错： Trying to copy blobs of different sizes.

Issue - State: open - Opened by huiyang865 about 7 years ago - 5 comments

#28 - 在VS2015+win10配置上，编译出现cundnn错误。

Issue - State: open - Opened by pengyulong about 7 years ago - 5 comments

GitHub / senlinuc/caffe_ocr issues and pull requests