NVIDIA Pretrained Models

Elements of Pretrained models.  Have not used this approach before, worth understanding its application. 

Jun 24, 2021

Fast-Track Production AI with Pretrained Models and Transfer Learning Toolkit 3.0  By Joanne Chang

Tags: Computer Vision & Machine Vision, DeepStream, featured, Healthcare & Life Sciences, Jarvis, Manufacturing, Metropolis, News, NLP / Conversational AI, Pretrained Models, Retail/Etail, Robotics, TLT, Triton

Today, NVIDIA announced new pretrained models and general availability of Transfer Learning Toolkit (TLT) 3.0, a core component of NVIDIA’s Train, Adapt, and Optimize (TAO) platform guided workflow for creating AI. The new release includes a variety of highly accurate and performant pretrained models in computer vision and conversational AI, as well as a set of powerful productivity features that boost AI development by up to 10x. 

As enterprises race to bring AI-enabled solutions to market, your competitiveness relies on access to the best development tools. The development journey to deploy custom, high-accuracy, and performant AI models in production can be treacherous for many engineering and research teams attempting to train with open-source models for AI product creation. NVIDIA offers high-quality, pretrained models and TLT to help reduce costs with large-scale data collection and labeling. It also eliminates the burden of training AI/ML models from scratch. New entrants to the computer vision and speech-enabled service market can now deploy production-class AI without a massive AI development team. 

Highlights of the new release include:

A pose-estimation model that supports real-time inference on edge with 9x faster inference performance than the OpenPose model. 

PeopleSemSegNet, a semantic segmentation network for people detection.

A variety of computer vision pretrained models in various industry use cases, such as license plate detection and recognition, heart rate monitoring, emotion recognition, facial landmarks, and more.

CitriNet, a new speech-recognition model that is trained on various proprietary domain-specific and open-source datasets.

A new Megatron Uncased model for Question Answering, plus many other pretrained models that support speech-to-text, named-entity recognition, punctuation, and text classification.

Training support on AWS, GCP, and Azure.

Out-of-the-box deployment on NVIDIA Triton and DeepStream SDK for vision AI, and NVIDIA Jarvis for conversational AI.  ... '