Anybody here who got CreateML object detection running?

The CreateML app wants annotation in the JSON format - neither the labels described in the beta documentation nor described in the WWDC videos are working, I always get scan errors. Is anybody here who got this working?

Replies

It's now running - but be aware to have the right JSON format, especially the right order for the coordinates:


[{"image":"sample.jpeg",

"annotations":[

{"label":"Label 2","coordinates":{"y":1132,"x":781,"width":413,"height":394}},

{"label":"Label 1","coordinates":{"y":1167,"x":308,"width":397,"height":375}}

]

},

....

]

Yes, JSON is flexible but not very tolerant to erroneous data.


Great you found a solution, and thanks for the feedback. Don't forget to close the thread on your own correct answer.

JSON above is not/no longer working...