Kibana 5.x - visualizing counts with nested? data: any guidance appreciated

Elastic Stack Kibana
1

Hello all,

I have been at this for a few hours now, seemingly simple task, and I am not sure of the best way to visualize the data we are working with and I was hoping I could reach out for help for guidance:

We have a list of images which will be identified as 'imageId:'

For each image, there are about 2,000 cells which we individually classify with neural network into one of ten possible classes. The goal is to allow users to drill down on specific images and visualize the distribution of cell classes in each image, similar to what you see with a histogram plot. Ideally, we would want to maintain each 'prediction' as a separate value rather than count the cells prior to indexing, as seen in the document below:

{   
"imageId": "000001.jpg",
"predictions": [
        "A",
        "B",
        "C",
        "A",
        "A",
        "A",
        "A",
        "C",
        "B",
        "A",
        "A",                         
        "A"
        ]
}

While the following is not ideal, we tried these as well:

{
  "imageId": "000001.jpg",
  "cellType": "A",
  "cellCount": 1,200
}
{
  "imageId": "000001.jpg",
  "cellType": "B",
  "cellCount": 100
}
{
  "imageId": "000001.jpg",
  "cellType": "C",
  "cellCount": 10
}

Also tried:

{
   "imageId": "000001.jpg",
   "A":1,200,
   "B":100,
   "C":10
}

Any help would be GREATLY appreciated, we've really been banging our heads against the wall. We must be missing something.

Many thanks,
T

2

The first format will be difficult in kibana until something like https://github.com/elastic/kibana/issues/2646 is available. I can't think of a way to make that work currently.

What does the second option look like using a bar chart with a terms aggregation on imageId, and a subbucket using terms on cellType, and the metric aggregation on cellCount?

3

Thank you very much for the reply. Here is what a bar graph looks like with option number two:

image.png1183×533 11 KB

So it looks like the aggregations you suggested were spot on. What I'm not sure of now is how to deal with all of the images being displayed in the bar graph until I click on the table to drill down on the image I want. Not sure it will be possible to get around that.

Greatly appreciate the help. Any other thoughts?

Thanks!
T

4

If we remove the top level terms aggregation on imageId, the bar graph will combine all images until a filter in the table is clicked, which may help or may be misleading. What would the ideal bar graph look like for you before an image id has been clicked on?

5

Ideally, I think that it would be empty.

6

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.