New Comp Vision Auto Table Detection...How to Make it Work?

[hellyars]

I'm trying to use this new fancy AIS Image Input to Image Template to Image to Text tool setup to auto detect a table(s).

No joy.  What am I doing wrong? Each page of the PDF is one big table.

UPDATE:  I believe you need to connect the output from the Image Template tool to the optional input of the Image to Text tool, which is not exactly how they explain it here: :Unlocking Insights from Images using Computer Vision 

There is a Markup field coming out of the Image Template tool.  But, the Image to Text tool does not allow me to select an image to process.

[mceleavey · Accepted solution]

Hi @hellyars ,

You simply select your image file from the input tool, then use the filter to isolate the one you want. Then you feed the image column into the "Image to Text" tool.:

M

[hellyars]

@mceleavey  I independently sorted it out, but thanks!

[dhouse]

@hellyars When the Image Input tool is connected to the Image Template tool, the Image Template tool goes into "Table detection mode".  The output of the Image Template tool is a template (similar to the manual template you would create if not in Table detection mode).  This template then needs to be fed into the Image to Text tool in the "T" anchor.  The Image input tool would then be connected to the "D" anchor of the Image to Text tool.  The Image to Text tool then uses the template to pull the tables out into pipe-delimited cells that you can parse further with traditional Alteryx tools.  See attached image, because that was a mouthful.

[Link86]

I am also having some difficulty. However, my pdf has two different tables on the first page and the rest just one table that is part of the table from the first page that I want. How do I just grab this one table, and then the rest of the tables on the pages. I have about 100 pages across 120 files that will eventually need to come through. Any advice would be helpful. Thank you.

[dhouse]

@Link86 I'm not quite sure I understand your use fully, but the output of the Image to Text Tool will be similar to the attached picture.  The page column corresponds to the page number in your pdf and the table0, table1, etc. columns represent the first, second, etc. tables extracted from that page.  You can then use the attached (handy) table parser macro to extract the  tables present in each cell in those columns and then process that data as you would any other tabular data.  

We would love to hear your thoughts about how we could make the output from the table extraction better if you have any suggestions.  

[ckelley0]

I followed the example you supplied and am getting some very inconsistent results.  Is there a way to tell the system how many rows and columns there are?