Challenge #95: Draw 1000 Apples in Under 60 Seconds

[JoeM]

Last week’s solution can be found here!

This week is Thanksgiving week in the USA and I was thinking of two ways we could take these challenges: 1) Make it super easy to match the short week or 2) Or long... long enough to give those of you balancing in-laws, distant relatives, and chaos this week a temporary escape. I hope the Weekly Challenge can be  that one work thing you really 'need to do today or my boss isn't going to be happy'.  I went for the latter. Weekly Challenge has got your back.

This week we are focused on working with data from the game of "Quick, Draw!". Here you can play a game where the computer asks you to draw a picture, and as you draw the picture live, the machine will guess what you are drawing based on a wealth of pictures from other users – machine learning at it’s finest!

This week we will try to take around one thousand pictures of apples from a special file type called .ndjson (new-line delimited JSON) generated from "Quick, Draw", reconstruct it into an image, and develop a heat plot of all the apples and some samples of apples that were drawn. Each record in the data set represents on drawing and contains an array of numbers that represent pixel coordinates on the canvas of the picture ranging from 0-255 for both x and y. The syntax of arrays are as follows:

[

  [  // First stroke

    [x0, x1, x2, x3, ...],

    [y0, y1, y2, y3, ...]

  ],

  [  // Second stroke

    [x0, x1, x2, x3, ...],

    [y0, y1, y2, y3, ...]

  ],

  ... // Additional strokes

]

For additional documentation on the simplified .ndjson file, see here.

The file we have provided contains 1000 apple drawings, but if you are interested in drawing 150K of them or something else entirely check out all the full library of data here.
As an output, try emulate the output that we made here in PDF:

Drawings will run onto a second page or more based on how it was sized

A tool from the predictive tool install was used in this solution. If you’d like to generate the visual shown in this sample PDF, please go to downloads.alteryx.com and download our predictive package. If not, just try drawing all the apples with a workflow!

This week there will be no start file! Instead we have attached two things - 1) The .ndjson file from which the drawing are constructed (we are not providing a start file because connecting to the data is a big part of the challenge) and 2) the PDF of the desired output.

[LordNeilLord]

Finally got there (or close enough for me)! I managed to get the heat map pretty quickly, the drawings took a little longer because the settings I needed were buried deep within some config settings. But compiling all of this into a report was a nightmare! A real nightmare! These reporting tools are awful, lets hope Plot.ly brings some usability to Alteryx reporting. I look forward to seeing the solution so I can see where I was going wrong.

Spoiler

[paul_houghton]

I found this task was both interesting and frustrating. The spatial part of the work was really interesting and intuitive. It stretched my knowledge and ability in the spatial tool set, I really enjoyed that portion. The Reporting portion of the task was just frustrating. The reporting tools are not intuitive, they not easy to use and they require far too many individual tools to achieve the final goal. The Visual Layout Tool is the first step on that path and hopefully D3 will add more to the visualising capabilities but there is a long way to go.

Spoiler

Completed workflow
Output
Full Output Image

[LilianH]

The crowd-sourced drawings are too cool, and I thought this was an interesting application of the heat map and spatial tools. Definitely had to play around with the reporting tools, and unfortunately everything got laggy once I brought in the report map tool. I couldn't get the apple drawings to appear as unconnected polylines without adding the macro, so I'm interested to see what others come up with!

Spoiler

Output

Workflow

[NicoleJohnson]

What a cool challenge!! Who knew Alteryx could draw?!? Very fun way to spend my Black Friday morning rather than battling the crowds at the mall...

Spoiler

Was actually able to draw the apples just fine, but took me some digging to figure out how to change the line color (because apples should be red, not grey) and then get them displayed correctly in a table. I'm on the same page as others about the reporting tools... they're not my favorite. BUT this definitely helped illustrate some of the cool, outside-the-box things you can do with the reporting & predictive tools, if you're willing to dig in a little deeper!


I tried it just for fun with one of the other NDJSON files too, the one for airplane drawings... pretty cool to see it work for a different drawing type as well!

Thanks for the awesome Thanksgiving week challenge, @JoeM! 

Cheers,

NJ

[patrick_digan]

Fun challenge!

Spoiler

I reused some lat long calculations that @Joe_Mako came up with in a previous challenge. My heat map was flipped from the solution pdf but I decided to leave it as is. I'm going to check out the other's solutions.

[estherb47]

Had some challenges with the mapping tool and getting the apples to render as more than blobs. Wondering if the tool just needs better documentation?

Anyways, my solution is attached.

[PhilipMannering]

Really enjoyed this one. I don't do enough image analysis.

Here's the output:

Spoiler

And here's what the workflow looks like:

Spoiler

[Natasha]

It took me quite some time to create points, but most of my time was spent on figuring out how to zoom to the drawing. I am getting used to charting tools more as a lot of challenges recently involved creating charts, but otherwise I am a happy Tableau user.

Fun challenge overall

Spoiler

[ggruccio]

Loved it!  I kept getting a stray apple that rendered huge, so forced to 25 columns and 40 rows.