VisualCortex – the Video Intelligence Platform connecting computer vision’s potential to real-world business outcomes – has adopted Weights & Biases‘ developer-first MLOps platform to systemize its end-to-end training process for Machine Learning (ML) models.
As foundational beta users of Weights & Biases recently-released ‘Launch’ module for automating ML workflows, VisualCortex’s ML team collaborated with Weights & Biases’ product team to help produce a scalable process for seamlessly managing and deploying training runs.
“Working closely with the Weights & Biases team, we’ve been able to provide feedback that’s been received with genuine thanks and then actually incorporated into their product – which is both rare and pleasing,” said VisualCortex Head of Machine Learning and Artificial Intelligence, Mike Seddon. “At VisualCortex, quickly delivering highly-accurate ML models, which can be reliably deployed into commercial environments, is a key differentiator for us and our clients in a competitive marketplace. To do that, we need to automate our training processes, easily determine which approaches are yielding the best results, and maximize the utilization of our ML training infrastructure at all times.”
“Weights & Biases is empowering us to efficiently execute and scale our training workloads, which enables us to get the most out of the hardware we use too,” said VisualCortex Co-Founder and CEO, Patrick Elliott. “We’ve integrated Weights & Biases into our core model training workflow as a source of truth and auditability. We can evaluate and compare many different model runs, perform sweeps to find data augmentations, and easily orchestrate disparate sets of compute nodes.
“We now have a robust ML model training process, which allows us to produce – and continually improve – highly accurate computer-vision-based insights for our clients. With Weights & Biases’ help, VisualCortex has the repeatable framework needed to bring video analytics into the mainstream.”
Weights & Biases VP of Product, Phil Gurbacki, said that gathering transparent product feedback from organizations training production-ready ML models was critical.
“Gathering input from professionals in the field – who are attempting to push the boundaries of ML model training, testing and deployment – is an important feedback loop for us,” said Gurbacki. “The team at Weights & Biases is proud to be helping VisualCortex develop and deliver market-leading models for people and vehicle-centric uses cases, including real-time object detection and action recognition, across a range of industries – from retail and commercial real estate, to car park management, smart cities and transport.”
Discover VisualCortex’s People Counting and Pedestrian Analysis Solution here:
Discover VisualCortex’s Vehicle Counting and Traffic Management Solution here:
Read the full VisualCortex – Weights & Biases case study here: