LogixStudio Gets Computer Vision Capability
Jacksonville, FL – LogixStudio, the core machine learning toolkit used by NLP Logix to rapidly develop and deploy predictive models has gained another capability: Computer Vision.
“The field of computer vision has seen tremendous advancements in the past 5 years,” said Matt Berseth, NLP Logix Lead Scientist. “These advancements enable technology to replace or augment tasks that are currently being done manually. Self-driving cars is a prime example, but others will be developing at a rapid pace. We expect computer vision technology to bring increased productivity through automation to our customers, starting with healthcare. To get ahead of this wave, we have invested considerable time and effort over the past year, to add computer vision capabilities to LogixStudio.”
To start the effort, Berseth chose to build the capability using a large image data set containing cancer tumors that were identified and labeled by experienced pathologists. He then customized a number of algorithms within LogixStudio, to be able to recognize the patterns within the image slides.
Now that a base set of image processing algorithms have been added to LogixStudio, the team at NLP Logix plans to further refine and improve upon the already impressive results. These refinements include adding additional code to the suite of proprietary model bindings that are contained within LogixStudio to integrate the models into a customers operating environment.
“The results Matt has been able to achieve are amazing,” said Robert Marsh, NLP Logix Chief Technology Officer. “Now we are making sure that the models can be integrated directly into a system that, for example, a pathologist may use to review specimen slides, giving them the power to have the machine read the slides right along with them.”
In addition to applications in healthcare, NLP Logix has already begun discussions with a large logistics company to apply the computer vision capability into their safety regimen. This includes taking and storing massive amounts of video data, data that up until now, was not being analyzed except in a reactive manner to an event.