神经网络训练 [英] neural network training
问题描述
如何为devnagarik手写检测进行神经网络训练。
请帮助我们这是最后一年的大学项目。
谢谢
sujitaparajuli
2013年5月24日编辑：
将问题移到作业部分。
- Leroy J. Gibbs
how to do neural network training for devnagarik handwriting detection.
Please help us this is for college project for final year.
thank you
sujitaparajuli
Edit on 24-MAY-2013:
Moved question into the homework section.
- Leroy J. Gibbs
推荐答案
A。请记住考虑提出问题的规则：
A. Please remember to consider the rules for asking a question:
- 你 搜索 或 Google搜索
获取解决方案？ - 具体！不要问我需要写预订申请。
准确指出您需要帮助的内容。 - 保持主题简短但具有描述性。例如如何更改对话框颜色？
- 尽可能将问题保持为简洁。如果您必须包含代码，请包含
最小代码片段 - 不要转储整个代码库。 - 正确标记您的问题。
- 您的问题可能会被其他人编辑或重新标记。任何不合适的内容都将被删除。
- 如果您有学校或大学的任务，假设您的老师或
讲师也在阅读这些论坛。 - 礼貌并且不要大喊大叫。这里的每个人都有帮助，因为他们喜欢帮助他人，而不是因为这是他们的工作。
- 如果其他人回复，请不要删除或清空消息。保持线程完好
并可供其他人搜索和阅读。 - 不要辱骂，冒犯，不恰当，骚扰董事会中的任何人或发布广告或垃圾邮件。这样做会让你开始并被禁止。 玩得好。
- Have you searched or Googled
for a solution? - Be specific! Don't ask "I need to write a booking application".
Specify exactly what it is you need help with. - Keep the subject brief but descriptive. eg "How do I change the dialog colour?"
- Keep the question as concise as possible. If you have to include code, include the
smallest snippet of code you can - do not dump your entire codebase. - Tag your question appropriately.
- Your question may be edited or retagged by others. Anything inappropriate will be
removed. - If you have a school or university assignment, assume that your teacher or
lecturer is also reading these forums. - Be courteous and DON'T SHOUT. Everyone here helps because they enjoy helping
others, not because it's their job. - Do not remove or empty a message if others have replied. Keep the thread intact
and available for others to search and read. - Do not be abusive, offensive, inappropriate,harass anyone on the boards or post
ads or spam. Doing so will get you kicked off and banned. Play nice.
2。看看这个：
http：//www.dfki .de /~sroa / papers / presentationgcnn.pdf [ ^ ]
我已将您的问题转移到作业部分，因为它既没有明确说明也没有具体也无法识别您放入在任何努力中你自己。
如果你做了进一步的研究，请发布另一个问题，并告诉我们你被困在哪里，为什么你被卡住了你有什么到目前为止。
欢呼，
Leroy
2. Have a look at this:
http://www.dfki.de/~sroa/papers/presentationgcnn.pdf[^]
I have moved your question to the homework section, since it is neither well-spelled out nor specific nor recognizeable that you have put in any effort yourself in yet.
Please post another question when you have done further reasearch and can tell us where you are stuck, why you are stuck and what you have done so far.
cheers,
Leroy
不容易。
诀窍是预处理你想要阅读的文本的扫描。我的建议：
1.提高对比度以尽可能多地消除噪音。您可以采取其他措施来提高准确性。
2.拍摄图像并拆分字符
3.标准化字符图像的大小。 1像素=> 1 NN输入。
4.单独处理每个角色。
步骤2相当困难，但我想这条线路一直在运行脚本的顶部将有所帮助，每个角色都是离散的。
显然你需要训练你的NN来识别角色。我建议使用BackPropagation网络，根据你能找到的每个角色的例子进行训练。你需要一个隐藏层（神经元的数量是试验和错误的问题），我建议每个Devnagari字符输出1个神经元。 这 [ ^ ]是训练BPN网络的一个很好的贯穿。
我发现有助于我的NN的两件事是在训练期间增加动量因子，在较小程度上一个boltzmann机器（你可以谷歌这两件事），虽然我的问题领域是完全不同所以YMMV。
Not easy.
The trick is to pre-process the scan of the text you want to "read". My suggestion:
1. Whip up the contrast to take out as much of the noise as possible. There may be other steps you can take to improve accuracy.
2. Take the image and split out the characters
3. Standardise the - size of the character image. 1 pixel => 1 NN input.
4. Process each character separately.
Step 2 is fairly hard, but I'd guess the line running along the top of the script will help, as will the fact that each character is discrete.
Obviously you'll need to train your NN to recognise the characters. I suggest a BackPropagation Network, trained against as many examples of each character you can find. You'll need a hidden layer (number of neurons is a matter of trial and error), and I suggest 1 output neuron per Devnagari character. This[^] is a good run-through of training BPN network.
Two things I found helpful for my NN was to add a momentum factor during training, and to a lesser extent a boltzmann machine (you can Google both these things), though my problem domain was quite different so YMMV.
这篇关于神经网络训练的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
