Computer Vision / Video Analytics

Inception Spotlight: AI Startup Neurala Sees 7X Speedup with NGC

To help businesses develop custom computer vision solutions quickly, Neurala, a member of NVIDIA’s start-up incubator Inception, has developed Brain Builder, a cloud platform that provides data scientists and developers that are new to deep learning with the ability to quickly and easily train neural networks. 

“Neural is an an end-to-end SaaS platform to streamline the creation of custom vision AI solutions,” the company wrote in a blog post. “By giving organizations an all-in-one tool for data tagging, training, deployment, and analysis, Brain Builder makes custom vision AI creation more efficient and accessible for AI beginners and experts alike,” they stated.

In the video below, the Neurala team describes how they improved performance in their workflow by using containers from NGC, NVIDIA’s hub for GPU-accelerated software.

NGC enabled the team to move from TensorFlow to a container version of TensorFlow on NGC with TensorRT integration, running on NVIDIA V100 GPUs for both training and inference, as well as testing other deep learning frameworks. 

Before NGC, the company’s training and inference times were simply too slow to provide value for their users. This shift to using containers on NGC resulted in a 7x speedup, the company said.

“To increase reliability, we now develop and test our networks on multiple frameworks,” Neves said. “In the past, that would have required a complicated and time-consuming installation. All the plumbing work required to develop Brain Builder would have hampered the delivery dates. Thanks to NGC, we were able to focus on building and delivering the solution on time.”

Speedup for the AI Annotator to learn (Training) or analyze (Inference) a single frame from a video after using NGC containers for Tensorflow on 1x V100 on AWS p3.2xlarge ec2 instance.

With Brain Builder running in NGC containers on NVIDIA GPUs, Neurala users can complete their work and build custom AI solutions significantly faster than would otherwise be possible, Neves explained.

Read more>

Discuss (0)