YCG09/chinese_ocr issues and pull requests

#377 - 和paddle resnet34 ctpn相比，哪个好？

Issue - State: open - Opened by nissansz about 1 year ago

#376 - potential bug in init.py

Issue - State: open - Opened by Alfusainey over 1 year ago

#375 - Dose it work on ARM?

Issue - State: open - Opened by sck-star over 1 year ago

#374 - 模型量化问题

Issue - State: open - Opened by yoummiegao almost 2 years ago

#373 - 想问一下为什么文本识别标签的txt文件夹中没有文本字符，只有坐标框，那gt在哪里

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

#372 - When run demo.py Segmentation fault (core dumped)

Issue - State: open - Opened by aiyodiulehuner about 3 years ago

#371 - 运行demo.py出现错误

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

#370 - 请问我使用你的数据集训练后，模型只会识别十个字符的图片。该如何解决？

Issue - State: open - Opened by MgArcher over 3 years ago - 1 comment

#369 - 无法加载VGG模型

Issue - State: closed - Opened by Jiangshuyi0V0 almost 4 years ago - 1 comment

#368 - 请问有谁碰到这个错误么Labels length is zero in batch 8[[{{node ctc/CTCLoss}}]]

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

#367 - How to understand the line-nms?

Issue - State: open - Opened by lhao0301 about 4 years ago

#366 - demo.py测试图片时，每张图片都很耗时，怎么解决。。。

Issue - State: open - Opened by lmw0320 about 4 years ago - 6 comments

#365 - 运行demo.py，第一次耗时较长，达到一百多秒，而且识别效果不佳。。这个是为什么？

Issue - State: closed - Opened by lmw0320 about 4 years ago - 1 comment

#364 - 无法正常执行demo.py,求指点

Issue - State: closed - Opened by lmw0320 about 4 years ago - 2 comments

#363 - 识别部分的一些使用心得

Issue - State: open - Opened by shining-love over 4 years ago - 5 comments

#362 - 5990个字符怎么会有重复的啊~

Issue - State: open - Opened by 15737939656 over 4 years ago

#361 - 请问自己构建数据集，如何实现“每个样本固定10个字符，字符随机截取自语料库中的句子”？

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

#360 - 文本行间距比较小的情况下文本检测会出现遗漏的情况，该怎么解决呢？

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

#359 - 继续训练的问题

Issue - State: open - Opened by shining-love over 4 years ago - 3 comments

#358 - 一个严重问题

Issue - State: open - Opened by shining-love over 4 years ago - 3 comments

#357 - GPU利用率很低

Issue - State: open - Opened by liangjing1 over 4 years ago - 1 comment

#356 - 数据集准备

Issue - State: open - Opened by 15737939656 over 4 years ago - 5 comments

#355 - loss变为nan是怎么回事呢

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

#354 - 怎么把densenet换成其他分类网络？

Issue - State: open - Opened by keithyh over 4 years ago - 1 comment

#353 - 可以使用不定长的图片进行训练么?

Issue - State: open - Opened by HongChow over 4 years ago - 2 comments

#352 - 重新训练loss值一直很高，acc一直为0的问题

Issue - State: open - Opened by doublegssc over 4 years ago - 3 comments

#351 - 尝试在win 10 上运行 demo.py

Issue - State: open - Opened by keithyh over 4 years ago - 3 comments

#350 - finetune 的效果很差

Issue - State: open - Opened by HongChow over 4 years ago - 5 comments

#349 - model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer='adam', metrics=['accuracy'])

Issue - State: open - Opened by zfxxfeng over 4 years ago - 1 comment

#348 - 相同大小的字，图片尺寸不同大小识别效果相差很大

Issue - State: open - Opened by 280868756 over 4 years ago - 8 comments

#347 - 准确率验证准确率都是1 什么情况但是用训练出来的模型结果却乱七八糟

Issue - State: open - Opened by Airflyctl over 4 years ago - 2 comments

#346 - data_train.txt中数据格式

Issue - State: open - Opened by keithyh over 4 years ago - 2 comments

#345 - 请问有人试过用efficientnet网络来做文字识别的么？

Issue - State: open - Opened by pingzi5233 over 4 years ago

#344 - densenet layers 层数问题

Issue - State: open - Opened by JJoohhnnn over 4 years ago - 2 comments

#343 - 在加载预训练模型VGG_imagenet.npy时，出错：IOError: Failed to interpret file '/data/pretrain_model/VGG_imagenet.npy' as a pickle

Issue - State: open - Opened by xjdx123 over 4 years ago - 3 comments

#342 - basemodel和model

Issue - State: open - Opened by carlcarl001001 over 4 years ago - 1 comment

#341 - 问什么识别速度那么慢？是配置不行吗？

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

#340 - cannot import name 'bbox'

Issue - State: closed - Opened by Ludasheng over 4 years ago - 2 comments

#339 - 有没有人能够分享下保存了结构，配置和权重的模型，楼主的模型文件好像只保存了权重。

Issue - State: open - Opened by an8852140 over 4 years ago

#338 - input_length 序列长度

Issue - State: open - Opened by EndSmart over 4 years ago - 1 comment

#337 - 请问Densenet是如何进行不定长字符识别的？

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

#336 - densenet+ctc和crnn+ctc哪个效果好?

Issue - State: open - Opened by swy0915 over 4 years ago - 2 comments

#327 - 加快訓練不定長圖片的速度

Issue - State: open - Opened by kellylin841015 over 4 years ago - 2 comments

#324 - chinese dataset

Issue - State: open - Opened by DayBreak-u almost 5 years ago - 1 comment

#322 - 只用來訓練數字時，不理想.

Issue - State: open - Opened by kaiyiuw almost 5 years ago - 1 comment

#318 - 运行demo.py时出现问题

Issue - State: closed - Opened by ghost almost 5 years ago

#311 - Synthetic_Chinese_String_Dataset的问题

Issue - State: open - Opened by swy0915 almost 5 years ago - 3 comments

#308 - DensNet训练数据集介绍

Issue - State: open - Opened by HJanefer almost 5 years ago - 3 comments

#306 - chinese_dataset

Issue - State: open - Opened by hujan01 almost 5 years ago - 1 comment

#302 - RGB图片训练模型，loss不收敛

Issue - State: open - Opened by CJ-XX2018 almost 5 years ago - 3 comments

#300 - tensorflow与keras的兼容问题

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

#299 - 训练自己数据集报错：tensorflow.python.framework.errors_impl.InvalidArgumentError: sequence_length(0) <= 159

Issue - State: open - Opened by cuimiao187561 about 5 years ago - 1 comment

#296 - 把识别模型.h5转到pb识别结果不一样

Issue - State: open - Opened by cuimiao187561 about 5 years ago - 2 comments

#295 - 如何将RGB图作为输入训练OCR识别模型？

Issue - State: open - Opened by CJ-XX2018 about 5 years ago - 1 comment

#287 - 花了很久時間，但Loss降不下去

Issue - State: open - Opened by kellylin841015 about 5 years ago - 4 comments

#280 - 请问怎么得到keras ocr识别后的准确度？

Issue - State: open - Opened by qicheng111 about 5 years ago - 2 comments

#278 - key.py

Issue - State: open - Opened by ql106 about 5 years ago - 1 comment

#273 - 模型训练的acc一直停留在85%， densenet网络的input的filters可以改成128来提高模型的拟合能力吗？

Issue - State: open - Opened by Asher-1 about 5 years ago - 4 comments

#270 - 训练时，每个batch的宽度不一样可以吗？

Issue - State: open - Opened by resetttttt about 5 years ago - 2 comments

#269 - 每个样本必须是10个字符吗？

Issue - State: open - Opened by jjl001 over 5 years ago - 13 comments

#268 - translate to english

Pull Request - State: closed - Opened by ajinkya933 over 5 years ago

#263 - 训练数据是否必须有良好语义？？

Issue - State: open - Opened by hello2sys over 5 years ago - 6 comments

#259 - ctpn/prepare_training_data/split_label.py 是否有问题？

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

#258 - 请问有谁试过在作者原模型上进行迁移学习的吗？

Issue - State: open - Opened by pingzi5233 over 5 years ago - 1 comment

#256 - 如何提升GPU的利用率？

Issue - State: open - Opened by Williamwsk over 5 years ago - 4 comments

#243 - CTPN训练数据图片和标签路径

Issue - State: open - Opened by KyrieYan1996 over 5 years ago - 3 comments

#241 - 使用小数据集fine-tune模型的问题

Issue - State: open - Opened by ericosmic over 5 years ago - 4 comments

#234 - 训练问题

Issue - State: closed - Opened by JJoohhnnn over 5 years ago - 9 comments

#226 - 急！densenet训练自己的数据（全英文字母和数字）为何loss和val_loss总是那么高呢,acc和val_acc保持相当，最高只能达到0.94，最低的时候也是1.2了，我逐个查看了数据，数据没什么问题，数据量为200w

Issue - State: open - Opened by zhangsn828 over 5 years ago - 6 comments

#223 - 在训练中报错：Labels length is zero in batch 14

Issue - State: open - Opened by ericosmic over 5 years ago - 8 comments

#219 - 关于数据集问题

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

#217 - 训练手写识别

Issue - State: open - Opened by wting861006 over 5 years ago - 2 comments

#209 - 用自己训练的densenet模型检测，识别不出汉字？

Issue - State: open - Opened by JJoohhnnn over 5 years ago - 22 comments

#198 - demo.py 加载densenet模型预测时报错，Attempting to use uninitialized value

Issue - State: open - Opened by bing1zhi2 over 5 years ago - 8 comments

#192 - 训练时的acc是行准确率还是字准确率啊

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

#185 - 测试的时候通过，训练的时候报了这个问题，也下载了最新的train.py，依然不行。python3.6

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

#181 - 求助,我在训练的时候遇到了问题,我使用的数据集是长度不相同的图片

Issue - State: open - Opened by 928234269 over 5 years ago - 2 comments

#177 - densenet网络结构的issue

Issue - State: open - Opened by xiaosi2017 over 5 years ago - 2 comments

#171 - 请教，我看别人只用DenseNet也能做ocr识别，为何还要加ctpn?

Issue - State: open - Opened by woshi91 over 5 years ago - 7 comments

#167 - 成功训练了ctpn和densenet，在测试过程中，文本识别结果显示没有空格，无法识别垂直文本。我怎样才能解决这个问题？

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

#165 - 如何修改batchsize大小，找了半天没找到

Issue - State: open - Opened by ziyuwzf almost 6 years ago - 5 comments

#164 - 测试时，模型加载问题

Issue - State: open - Opened by mayidu almost 6 years ago - 5 comments

#162 - acc为0 咨询!

Issue - State: open - Opened by shenxiao2019 almost 6 years ago - 11 comments

#161 - 迁移学习

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

#147 - 运行demo.py用时9秒，正常吗？

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

#145 - 关于修改char_std_5990.txt，再重新训练的问题

Issue - State: open - Opened by Gretaing almost 6 years ago - 5 comments

#141 - 在处理训练数据的时候label存在问题

Issue - State: open - Opened by xxlxx1 almost 6 years ago - 12 comments

#139 - 运行train报错

Issue - State: open - Opened by lianqingsong almost 6 years ago - 29 comments

#131 - 测试发现，只要是白色的字，识别不理想，转成黑色字能识别，有些疑问：1. 作者是基于灰度图像训练的，数据集原因还是灰度图的效果优于rbg图像？2.如果使用彩色rgb训练？将输入batchsizehw*c的channels改成3就可以？后续的网络结构无需修改？谢谢作者的分享

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

#127 - 关于训练数据的问题？

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

#89 - input_length[i] = imagesize[1] // 8

Issue - State: open - Opened by mocun172 about 6 years ago - 2 comments

#86 - densenet单独测试效果不佳

Issue - State: open - Opened by 18813055625 about 6 years ago - 5 comments

#85 - 请教一下，如何识别空格字符信息？

Issue - State: open - Opened by fcqfcq about 6 years ago - 9 comments

#84 - 识别类似食品经营许可证这种图片，在服务器上（有一块gpu）还要9秒钟

Issue - State: open - Opened by MRCHEN3303 about 6 years ago - 3 comments

GitHub / YCG09/chinese_ocr issues and pull requests