Hot Dog or Not Hot Dog: Build a very simple machine learning app with Bubble & Watson

Hey Bubblers!

I wrote a post explaining how to build a very simple Machine Learning app with Bubble + Watson Visual Recognition:

Hot Dog or Not Hot Dog

This is intended for Bubble newbies, so may be too simple for a lot of the forum regulars but hopefully some of you find it useful (especially those not familiar with the API Connector)!



That looks like a very good exercise. I’ll try it out and let you know. :wink:

It’s great- but I can’t get it to work - nice job in being clear though, but I think IBM changed the login process? anyway any suggestions would be very helpful…

Hi @lauren - where specifically are you running into trouble? Happy to try to help.

Hey @lauren, IBM made indeed some changes. I was confronted with the same difficulties when I tried to follow the tutorial.

You can check the editor of this test app (the API key has been removed, so running the app itself won’t show you much, only the editor will be helpful): test-ai | Bubble Editor

Here below is the detailed explanation I obtained from Watson’s very supportive customer support team (same day reply by email for a non-paying customer).

Watson services migrated to token-based Identity and Access Management (IAM) authentication. Visual Recognition instances created after May 22nd (2018) use this method, which in turn changes the way the api keys are identified in the api calls as shown in the example below:

curl -X POST -u "apikey:{your_api_key}" --form "images_file=@fruitbowl.jpg" \

In addition, you can check the following documents for detailed information about these changes:
IBM Cloud Docs
IBM Cloud Docs

@vlad: I had a lot of fun with your tutorial, and built a derivative concept based on it (basically the same thing but allowing users to change the search term, and see the identified labels/classes by both IBM’s Watson and Google’s Cloud Vision API). Thanks a lot!

Hi @vlad I can’t find the tutorial for Hot Dog or Not Hot Dog. When I click on your link it goes to your blog.

The post is towards the bottom of the blog entries page, 11th down at the mo.

Looks like maybe the original link changed ID of the post, hence landing on the blog list. Anyhow, great post Vlad!

1 Like

Thanks @luke2 I see it now

Vlad - okay, count to 10… with your and Lucien’s help I got the API working… but when I’m trying to add the condition… 'Result of step 1…" - I don’t know how to do that … I don’t see ‘result of prior step’ or anything like it… sorry to be a pain, but love a hint


Hi @lauren! First of all thanks for flagging the issue and sorry for the delay. I have it on my list to update the blog to reflect Watson’s latest changes but just haven’t gotten to it yet, hope to do it soon.

As far as the “Result of step 1” issue - do you mind pasting a screenshot of your workflows here? Happy to take a look. You can also add as a collaborator instead and I’ll look directly within the app.

And @Lucien- thanks for the update from IBM, super helpful. Glad you enjoyed the tutorial and awesome that you were able to build a similar concept!

Finally I can’t seem to edit the original post but here’s the updated link directly to the blog post (we did indeed change it)

Thanks for the help Vlad - I got it working by staring at your example, realizing (new user mistake) that I had to do a bit of setting up data types that weren’t shown in your tutorial - things like making sure the API call to google worked and had the right classes identified… thank you it’s fun to see it work.

@vlad Thanks a million man! You are the reason why i have started bubbling.

1 Like

How did I miss this post?! This was such a great episode on Silicon Valley. FYI, NSFW due to language. :slight_smile: