Streamlined Compatibility Between Zetane and Keras Provides Unprecedented Transparency to Machine Learning Models
Press release: AI professionals gain frictionless access to detailed visuals of most machine learning projects.
[Montreal, Canada; May 27, 2021] A major software release announced today by Zetane Systems will enable a greater pool of professionals in the artificial intelligence (AI) sector with frictionless access to detailed visuals of machine learning projects. The enhanced compatibility provides greater interoperability between software products from the Montreal-headquartered start-up with models made with the popular open-source Keras deep-learning API of the TensorFlow platform.
Keras is a software platform of choice for machine learning and data science professionals used to develop most industrial applications of machine learning. Software tools by Zetane — the Viewer and Engine — provide the most extensive displays of machine learning models and their processing of data. Such detailed and intuitive visuals of AI projects provide methods to better understand the inner workings of complex artificial neural networks. In turn, greater understanding and transparency offer new strategies to audit models and reduce time-consuming guesswork typically required for optimizing and testing AI products. The merging of capacities between both software platforms centres on enabling professionals to load and run native Keras files within Zetane, thus replacing a complex process of scripting computer code and employing a Python API to convert Keras files into compatible formats.
The strengthened compatibility between Zetane and Keras will provide several benefits due to this gain in efficiency and transparency. Of primary import is the increased ease by which professionals can collaborate and conduct visual inspections of projects completed by colleagues and external businesses. By removing the need to convert computer files to a common format, professionals now can share and open native Keras files and begin inspecting and testing the performance of a machine learning project in the Zetane Engine using an intuitive graphical user interface. Efforts to validate the proper functioning of an unfamiliar model by having peers test its processing of their chosen data inputs are now simple and fast, requiring only common data inputs in the form of digital images, NumPy files or protobuf files.
“Achieving this latest product milestone goes beyond bundling Keras with our cutting-edge visualizer and development tools. We’re bringing depictions of artificial neural networks into the modern era of user interfaces required by enterprise software that meets the standards for user-friendliness demanded by astute consumers,” said Jonathan Magoon, Chief Product Officer and co-founder of Zetane.
Moving towards shared industrial standards for AI
Early phases of new technologies typically pass through a “wild West” period where shared technology standards have yet to be established. AI is no exception. Though a plethora of cutting-edge development tools for AI have flooded the sector as of late, many use distinct file formats and operate with incompatible interfaces.
As the AI sector matures, so too does the establishment of common industry standards. The global standards set by the consortium, Open Neural Network eXchange (ONNX) community of the Linux Foundation, is one notable effort. This community of leading technology companies — which includes Amazon Facebook, IBM and Zetane — promote common standards for representing machine learning models as a means to lower barriers to AI innovation. Of note is the Zetane Viewer recently being classified as the official visualization software tool for the ONNX Community. This latest compatibility enhancement with Keras represents the company’s next step in continuing its leadership role in promoting open-source standards and access to free visualization tools throughout the AI sector.
About Zetane Systems
Headquartered in Montreal, Zetane provides the software and services required to de-risk and trust the deployment of AI innovations in industry. Rather than work with abstract computer code, our software provides a detailed visual representation of AI solutions to bridge the gap between AI experts and industry experts in order to support proper testing, validation, debugging, optimization and monitoring of AI. AI experts gain access to many easy-to-use features that save time and eliminate guesswork, and domain experts without expertise in AI can contribute and conduct needed performance and safety audits. The Zetane Viewer is an official visualization software tool for the ONNX.ai Community of the Linux Foundation, available free to download at zetane.com/gallery. Follow @ZetaneSystems for updates.
