Biarca has significant experience in the design, training, optimization, and production use of industrial-type machine vision systems, using best of breed open source image processing and visual classifier toolkits such as OpenCV and OpenCL. We enhance the performance of these systems by scaling them out on cloud platforms using open source clustering frameworks.
Realtime and Predictive Analytics
Biarca’s success in improving throughput on individual classifiers and in breaking down machine learning tasks into scalable units of work that can be performed on cloud substrates has led to an increasing number of engagements to create realtime predictive analytics solutions: classifiers whose predictions are used immediately for “fuzzy” decision support. Such systems find frequent application in marketing, business intelligence, financial market research and many other use-cases, where high accuracy of individual predictions is less important than speed and rough-and-ready, “ballpark” accuracy.
Reinforcing its work in machine vision and realtime analytics, Biarca collaborates with customers to develop good classifiers and training datasets for different problem domains. Much of this work is done with TensorFlow and similar tools.
Case Study: Cloud Migration with Machine Vision
Biarca recently helped a global manufacturer migrate their manufacturing control software from one public cloud provider’s VM infrastructure to a Biarca-managed, highly-available container-oriented solution running on another public cloud. The resulting solution was less expensive and significantly more resilient than what it replaced. Biarca was also able to leverage analytics and machine learning facilities available on the host platform to improve classification of visual inspection imagery. The result was to increase manufacturing quality.