H2O Hydrogen Torch
Democratize State-of-the-Art Deep Learning
2022 AI Tech Awards | H2O Hydrogen Torch awarded “Best in Deep Learning Technology”
-
No Code Deep Learning
-
Signature Detection
-
Manufacturing Product Defect Detection
Marketing
Hotel Recommendations
Financial Services
Document Signature Verification
AI4Good
Deforestation Segmentation
How Data Scientists use H2O Hydrogen Torch
Image
Classification and Regression
Object Detection
Semantic Segmentation
Metric Learning
3D Image
Classification and Regression
Semantic Segmentation
Text
Classification and Regression
Token Classification
Span prediction
Sequence to Sequence
Metric Learning
Video
Live stream object detection
Video Segmentation
Audio
Classification and regression
Speech to text
Benefits
Democratizes Deep Learning
Train state-of-the-art models with no coding needed
Deploy models easily to external Python environments or directly to H2O MLOps
Speed
Quickly experiment with hyperparameters to tune model performance
Ready-to-use library of problem types that continuously expands based on Data Science trends
Accuracy
Train deep learning models using training techniques developed by Kaggle Grandmasters
Achieve results by using the best neural network architectures and transfer learning
How-it-Works
Set up and manage deep learning experiments to maximize your model development.
Choose model architecture and most relevant tuning parameters
Quick start options provide the most important settings
Expert settings provide a wide range of training approaches
Select the training set and problem type within H2O Hydrogen Torch, and it will automatically learn the data and create models
AI-ML engine that uses multiple computer vision and NLP algorithms for diverse AI tasks (eg, entity recognition, grouping and set identification)
Optimize neural network and perform hyperparameter tuning without the need to change any code
Inspect and analyze your models within the tool.
Track progress during model training, assessing model accuracy and analyzing predictions of the model
Reveal errors in the data that supply ideas on how to improve the model
H2O Hydrogen Torch packages all the models automatically and provides a format which can be published to:
MLOPS, which is part of the H2O AI Hybrid Cloud
Your cloud or on-premises environment of choice
Apply deep learning models made with H2O Hydrogen Torch across a variety of business problems by:
Developing use-case specific apps built with H2O Wave
Importing into existing applications via APIs.
Capabilities
Prediction Insights that give Data Scientists ideas for how to improve models for next iterations.
Streamlined model deployment native to the H2O AI Cloud that can also be imported to other model serving platforms
Comprehensive set of techniques for deep learning model tuning selected from the best practices of Kaggle Grandmasters
Meets novice Data Scientists where they are with a no-code UI and fosters growth opportunities through practical experience and documentation
Multi GPU training support
Supported Deep Learning Problem Types
Text Data and Natural Language Processing (NLP)
Text Classification and Regression
Predict customer satisfaction from transcribed phone calls and categorize incoming emails requesting customer support to send to relevant departments
Token Classification
Extract entities such as disease names from medical text with multi-lingual support
Span Prediction
Find relevant information from medical transcripts and build a company-specific question-answering system
Sequence-to-sequence
Simplify text that contains domain-specific terms and summarize text for better understanding of its content
Metric Learning
Detect fake reviews that display similarities and find similar questions in user forums
Image, 3D Image and Video Data
Image Classification and Regression
Predict images for abnormalities in medical X-rays and classify types of landscapes from drone photos.
Object Detection
Detect vehicles from traffic and drone cameras and find objects required for assembly line processes.
Semantic Segmentation
Locate objects of interest in medical images, segment roads from video dashcams and remove objects from the background
Instance Segmentation
Detect and delineate distinct objects of interest in biological images, segment individual cars, pedestrians, or other objects from a video captured using a dashcam
Metric Learning
Find similar images to those in a training database, such as unauthorized use of a brand logo or duplicate items on a retail website
Featured Success Story
H2O Hydrogen Torch has been a key enabler in helping us operationalize machine learning for shifting data. We can get from a new dataset to a deployed model and updated tables in our data warehouse in a couple of days instead of weeks.”
CTO, Aura.ceo
Learn More
H2O Hydrogen Torch is available in the H2O AI Cloud. Request for an H2O team member to walk you through a demo.