Showing results for 
Search instead for 
Did you mean: 

Visual AI

Community Team
Community Team
6 4 6,291

(Part of a model building learning session series.)

You've built your first set of image models using DataRobot's new Visual AI capability. So, now what comes next? How can you improve model accuracy? How can you understand what the models are learning? And, all of the built-in Image Insights look cool—how can you use them to drive modeling decisions?

Join Ivan Pyzow from the DataRobot Visual AI team as he explains how to:

  • Identify overfitting and underfitting using Activation Maps.
  • Use Image Embeddings to identify target leakage.
  • Identify use cases for sorting your images via bootstrap labeling with DataRobot.
  • Advance-tune more accurate models by leveraging domain knowledge.

This session will keep the learning practical by walking through a number of projects in manufacturing, agriculture, retail, and home insurance, with takeaways that will be applicable to any organization and use case.


  • Ivan Pyzow (DataRobot, Deep Learning Engineer)
  • Jack Jablonski (DataRobot, AI Success Manager)

More Information

DataRobot Community: 

If you’re a licensed DataRobot customer, search the in-app Platform Documentation for Visual AI Overview and Using Visual AI.

Let us know what you think!

Have questions not answered during the learning session? What to continue your conversation with Ivan? Post Your Comment here or send email to We're looking forward to hearing from you!

Data Scientist
Data Scientist

That was such a great walk through! Nice work team!

Lolita Celsi
Computer Board

As a novice to AI, I'm happy to have an introduction to today's tools for data analysis and applications.

DataRobot Employee
DataRobot Employee

Hi @Lolita Celsi! Thank you so much for attending our Visual AI Learning Session, it's really great to hear that you found the session valuable. Please keep your comments coming!

Image Sensor

It's very interesting at 26:41 that said : build a simple classifier to exclude those . sounds like to build another classifier that excludes those mistakes caused by barcode appear in those pictures. Is that a DataRobot's function for everybody or advanced engineers' special technique ?  Point me to an article would be good enough if they exist. Thank you! 


26:39 resample our data maybe as roger saying

26:41 build a simple classifier to exclude those um so just

26:45 to highlight our observations and actions  


Welcome to the DataRobot Community!
We're so happy you're here! If you haven't already, check out helpful tips for navigating through the community, as well as guidelines to keep in mind while you're sharing and learning along with the rest of us. Have a look at the latest community newsletter.

Just getting started using the AI Cloud Platform Trial? Have a look at these basics and FAQs.

DataRobot Release 7.2 Is Here!
Learn about changes in DataRobot Release 7.2. You can learn more about AI Cloud and DataRobot 7.2 in the recent AI Cloud Launch Event. And, if you have questions about the release, ask them right here!