606120_en-US

liliconcom

Classification Problems - handwritten document to be recognized

Hi all, 

speaking with couple of customers we noticed a problem related to import handwritten notes into JMP.

It’s often notes made by operators or certificate of analysis for feedstock  or sheets that they receive from suppliers and mostly numeric data. In an attempt to explain properly let me send you a link we have for the internet to show you what it can look like: (45) AI Learns the Numbers - YouTube. The interesting think is that Python code (if I'm not wrong) is available here: Copy of AI Learns the Numbers Fourth Model.ipynb - Colaboratory (google.com)

 

I watched also a Discovery Summit talk from 2019 on the same topic: Solving Classification Problems With Deep Learning ( 2019-EU-30MP-100 ) - JMP User Community

 

Was wondering if you know an add-in already in place to do that or you have any idea to take advantage of the open code available.

Please ping me for extra details if interested! 

Thanks 

 

Marco 

 

Re: Classification Problems - handwritten document to be recognized

There hasn't really been anything native in JMP or in Add-in's that I'm aware of to handle image analysis or OCR.  That's generally not been a space we've gone after - image analysis.  You might get some traction with t-SNE or U-MAP on classifying things (Analyze > Multivariate Embedding).  The standard for doing this image work these days is generally considered GNN's - not something we do.  

Re: Classification Problems - handwritten document to be recognized

thanks @MikeD_Anderson.

I would like to find a solution for these customers so I will further investigate the problem.