H2O.ai Democratizes State-of-the-Art Deep Learning for Data Scientists and Developers with H2O Hydrogen Torch

Application development that combines sophisticated technologies such as AI and ML is progressing, this time by combining multiple deployment outcomes into a single general-purpose, no-code platform. Line-of-business users may easily transition from studying data records to natural language processing, image, and video outputs in this manner.

Because so many software developers can only focus on one of those outputs at a time, this type of adaptability for non-IT personnel hasn’t been offered on the market. Market heavyweights like DataRobot, Amazon Web Services, Microsoft, DataBricks, and SAS don’t provide this particular feature. H2O.ai, on the other hand, has set out to overcome this problem.

The company is situated in Mountain View, California. H2O.ai has announced H2O Hydrogen Torch, a significant new addition to its open-source platform. This feature is a deep-learning training engine that, according to the company, makes it easy for every size and area of the company to generate a cutting-edge image, video, and natural language processing (NLP) models without scripting. These models may be utilized in the field to uncover fresh business insights about consumers, rivals, the market, and other topics.

H2O Hydrogen Torch was created by Kaggle Grandmasters, the world’s most outstanding data scientist, and the solution handles the problematic elements of developing world-class deep learning models automatically, according to CEO and cofounder Sri Ambati. Until now, coding and tuning accurate deep-learning models necessitated substantial knowledge and effort. Because data scientists are among the highest-paid IT workers, these expenditures may be costly.

Smart LOB employees, data scientists, and developers, according to Ambati, can quickly create models for a variety of image, video, and NLP processing use cases, such as identifying or classifying objects in images and video, analyzing sentiment, or finding relevant information in texts, using a simple, no-code user interface.

Monitoring foot movement in public buildings, malls, and businesses, for example, and noting the frequency of visitors and where they go from place to location are examples of video use cases. The platform may be used by retailers to see which sales displays are the most popular. According to the CEO, all of the data is instantaneously gathered and accessible for queries in the H2O.ai analytics engine.

“There’s a lot of unstructured data out there,” CEO added, referring to photographs and text in businesses. “There’s a lot of promise that hasn’t been realized yet.” The purpose is to let and empower people to create state-of-the-art models for many sorts of use cases. We’re basically offering them these features in Hydrogen Torch to solve different kinds of use cases.”

According to several analyst estimates, unstructured data accounts for 80 percent to 90 percent of data, but only a tiny fraction of firms can get value from it, according to the CEO.

Deep-learning models have the potential to transform industries such as health care, which can use computer-aided disease detection or diagnosis based on medical images, insurance, which can automate claims and damage analysis based on reports and images; and manufacturing, which can use predictive maintenance based on images, video, and other sensor data, according to Ambati.

Image and Video Processing

Hydrogen Torch may be taught for classification, regression, object identification, semantic segmentation, and metric learning on pictures and videos, according to the CEO. Hydrogen Torch, for example, can evaluate medical X-ray pictures for anomalies in a medical context, with a “person in the loop” making the ultimate judgment. According to the CEO, object detection at a manufacturing facility to assess whether a part is missing, or metric learning to alert an online store to duplicate photos on a website, are two more image-based use cases.

Natural Language Processing

Hydrogen Torch may be taught for text classification, regression, token classification, span prediction, sequence-to-sequence analysis, and metric learning for text-based or NLP use cases. Natural language processing has many applications, from estimating consumer happiness from transcribed phone calls to using sequence-to-sequence analysis to summarise a massive amount of text, such as medical transcripts.

According to the CEO, these models may then be packed automatically for deployment to external Python environments or straight to H2O MLops in a consumable manner for production.

More than 20,000 worldwide enterprises, including AT&T, Allergan, CapitalOne, Commonwealth Bank of Australia, GlaxoSmithKline, Hitachi, Kaiser Permanente, Procter & Gamble, PayPal, PwC, Unilever, and Walgreens, utilize H2O.ai’s platform, which presently provides a free trial, according to the CEO.

References:

  • https://www.h2o.ai/ai-cloud-test/news/press-releases/h2o-ai-democratizes-deep-learning-with-h2o-hydrogen-torch/
  • https://h2o.ai/platform/h2o-hydrogen-torch/
  • https://www.bloomberg.com/press-releases/2022-02-17/h2o-ai-democratizes-deep-learning-with-h2o-hydrogen-torch
  • https://venturebeat.com/2022/02/17/h2o-ai-democratizes-deep-learning-for-companies-of-all-sizes-with-hydrogen-torch/