Introducing Visual AI for DataRobot Automated Machine Learning

April 2, 2020
· 6 min read

Visual AI is New in DataRobot 6.0

In Release 6.0 of DataRobot, we are thrilled to announce a ground-breaking new capability in our Automated Machine Learning product. DataRobot Visual AI gives you the ability to easily incorporate image data into your machine learning models alongside tabular and text-based data types. This enables your organization to get value from computer vision, right away – all with the same DataRobot usability, workflow, visuals, and other UI features you know and love.

DataRobot Visual AI interfaceClick image to see the full screenshot.

Analysts and data scientists of all skill levels can now drag and drop images into DataRobot to prepare, build, and deploy highly accurate deep learning models, in a fraction of the time compared to alternative tools. You can get started with just a few hundred images, so models can be trained in minutes or hours (not days), and the best part is, you don’t have to purchase expensive Graphical Processing Units (GPUs). DataRobot Visual AI is optimized to perform best on the hardware you already have in place.

With Visual AI, you can focus on solving business problems with image data without having to worry about gaining deep learning skills and spending money on infrastructure before you’ve built your first model.

How Do Organizations Use Visual AI?

Visual AI can be used across all industries in a wide variety of use cases. Retailers can use computer vision to improve the customer experience, detect when a product is out-of-stock on store shelves, or even watch for suspicious activity to help with loss prevention. Manufacturers can use Visual AI to identify product defects in real-time. As the parts and components come off their production line, images can be fed into their model to flag potential defects and avoid problems further downstream.

Insurance companies can conduct more consistent and accurate vehicle damage assessments to help reduce fraud and streamline the claims process. Healthcare providers can use image-based neural networks to automate the examination and diagnosis of health issues from MRI’s, CAT scans, and X-rays.

classifying-disease-with-visual-AIAbove: Classifying blood diseases with Visual AI. Click image to see the full screenshot.

During the recent Visual AI private beta program, our customers had a varying number of problems they were able to tackle. From using images of gas stations to help better plan where to focus marketing spend, to the automated labeling of apparel from fashion photography for a leading eCommerce website.

Images. But Not Just Images

Visual AI allows you to build binary and multiclass classification and regression models with images. You can use it to build completely new image-based models, for example, to detect defects in steel or classify diseases in plants. You can also use it to add images as new features to existing models. This helps improve model accuracy from the fresh perspectives that images provide. For example, a hospital readmissions model built on tabular data, with features such as diagnosis, age, and gender can be enhanced with more diverse information such as surgeon notes, and with Visual AI, images from the patient’s MRI.

featured1Above: DataRobot model Blueprints handle diverse types of data. Click image to see the full screenshot.

With Visual AI, we have shifted the market for image-based machine learning and computer vision. Now anyone using our Automated Machine Learning product can get value from their image data. Moreover, the ability to blend diverse types of data together in a single model is truly unique to DataRobot and is a game changer. We can’t wait to see how our customers take advantage of this to innovate in completely new ways.

Just How Easy is Visual AI to Use? 

In a word – simple. This is what you do in 10 easy steps:

1. Create your zip file.

Put your images into different folders to classify them or use a csv file if you want to add lots of additional features. Then zip everything up (check out our Community article that describes this process in more detail).

2. Drag-and-drop the zip file into DataRobot.

Either drag your images straight into a new project or upload the zip file into DataRobot’s AI Catalog to share the image dataset with others.

3. Pick your target and hit Start.

Just like you do with a non-image project today. It’s literally that easy.

4. Explore your image dataset.

Our automated EDA will show you lots of interesting statistics about your dataset features, including the identification of missing and duplicate images.

explore-the-datasetClick image to see the full screenshot
5. Sit back and relax.

DataRobot will automatically select, train, test and compare a whole host of cutting-edge deep learning algorithms then recommend the best one for your use case. This will take a few minutes or maybe a few hours, but not too long.

6. Evaluate the recommended model.
Once DataRobot has built and recommended a model, you can evaluate its accuracy using our familiar automated visualizations.

We also added some new ones specifically for images. These include the Neural Network Visualizer, Image Embeddings and Activation Maps.


They show you exactly how the model pre-processed the data, the algorithms it selected and even where the neural network looked in the image for every single prediction.

accuracy-parametersClick any of the images to see the full screenshot
7. Tune and tweak (if that’s your thing).

If you want to refine your model before you deploy it, we expose all of the models advanced hyperparameters for you to play with.

advanced-tuningClick image to see the full screenshot
8. Click to deploy.

Congratulations! With a single click, you just deployed your image model into your production environment.

DeploymentClick image to see the full screenshot
9. Monitor, manage, and evolve.

We provide all the tools you need to view the status of your model in real-time.  You can also see how it performs over time and which features have drifted. When you need to switch out your model with a new version, you can do that with absolutely no service interruption.

featured4Click image to see the full screenshot
10. Go buy a cape. You’re a deep-learning super hero, you should wear one.

Wait…What? No GPUs?!

Yes, you heard that correctly. To get started with Visual AI you just use the DataRobot environment you already have in place today. You don’t need to go out and purchase a bunch of expensive GPUs. This is because DataRobot Visual AI ships with pre-trained neural networks that can be used to build your model in a fraction of the time, and with a much smaller volume of images, than training a neural network from scratch. What’s more, right out of the box, accuracy is more than comparable, if not better than many models trained from the ground up.

In a recent test we used Visual AI to build a multiclass model based on ~14K images of natural scenes around the world (e.g. buildings, forests, streets, mountains, etc) from a competition published by Intel on Analytics Vidhya. We were staggered with the results. Visual AI trained roughly 40 models in just over two hours using commodity hardware only (i.e. no GPUs). Our accuracy right out of the box was just over 92% and with a little tuning we added another couple of percentage points. Not only were these results impressive in terms of accuracy and performance, but the fact that this test was performed by a business analyst with minimal data science experience and absolutely no deep learning skills at all, is simply amazing.

Confusion-MatrixAbove: DataRobot’s Confusion Matrix shows high model accuracy across all classes. Click image to see the full screenshot

Get Your Hands on Visual AI Today

Visual AI is part of DataRobot 6.0. If you’re an existing DataRobot Automated Machine Learning customer, you need to contact our customer support team to request the feature be enabled for your account. It’s available in our Managed solution today, and if you are running DataRobot v6.0 on-premise or in a private cloud, your DataRobot account team will help you enable it.

Build models with Visual AI right now using DataRobot Automated Machine Learning on the infrastructure you already have in place. It’s easy to use, so you don’t need deep learning expertise, and it gives you accurate models that are easy to understand. So you can explain every single prediction to those that need to know. When you’re ready, you can deploy your models with a single click, and then monitor and manage them to keep them accurate. You can even update them without interrupting service.

At DataRobot we are proud to bring Visual AI to the market. It is truly unique in the field of automated computer vision and unlocks new use cases, never before possible as well as improving your existing models. So what are you waiting for? Take your AI strategy to the next level with Visual AI today.

More Information on Visual AI

You can also visit the DataRobot Community to learn more about DataRobot 6.0 and watch a demo of Visual AI.

On-demand webinar
DataRobot Visual AI: See the Big Picture with Automated Computer Vision
Watch Now
About the author

Value-Driven AI

DataRobot is the leader in Value-Driven AI – a unique and collaborative approach to AI that combines our open AI platform, deep AI expertise and broad use-case implementation to improve how customers run, grow and optimize their business. The DataRobot AI Platform is the only complete AI lifecycle platform that interoperates with your existing investments in data, applications and business processes, and can be deployed on-prem or in any cloud environment. DataRobot and our partners have a decade of world-class AI expertise collaborating with AI teams (data scientists, business and IT), removing common blockers and developing best practices to successfully navigate projects that result in faster time to value, increased revenue and reduced costs. DataRobot customers include 40% of the Fortune 50, 8 of top 10 US banks, 7 of the top 10 pharmaceutical companies, 7 of the top 10 telcos, 5 of top 10 global manufacturers.

Meet DataRobot
  • Listen to the blog
  • Share this post
    Subscribe to DataRobot Blog
    Newsletter Subscription
    Subscribe to our Blog