AI developer Deci announces discovery of efficient computer vision models

July 14, 2021
The new DeciNets family of image classification models are designed to require less computing power than established technologies.

Deep learning software developer Deci, based in Tel Aviv, Israel, announced today the discovery of a new family of image classification models, called DeciNets, that require less computing power than established technologies.

Automated Neural Architecture Construction (AutoNAC) technology enables developers to automatically design and build deep learning models that according to Deci provide superior performance than other architectures such as Google-scale Neural Architecture Search technologies. AutoNAC is hardware-aware, allowing the technology to optimize model performance for cloud, edge, and mobile computing environments.

Deci reports testing AutoNAC via optimizing deep learning models using the NIVIDIA T4 and Jetson Xavier NX edge GPUs. Image classification model testing with the ImageNet benchmark dataset and using AutoNAC technology led to the discovery of the DeciNet models.

AutoNAC technology has previously been used to accelerate the inference speed of the ResNet050 neural network, via collaboration with Intel. To learn more about DeciNets, visit the Deci website.

Share your vision-related news by contacting Chris Mc Loone, Editor in Chief, Vision Systems Design


Voice Your Opinion

To join the conversation, and become an exclusive member of Vision Systems Design, create an account today!