Cogniac is a platform for visual task automation. In manufacturing, tasks can be prone to human error. It’s estimated that 23% of all unplanned downtime in the industry arises from it, a rate that is reduced to 9% in other segments.
The platform connects machine vision cameras, drones, security cameras, smartphones, and other sources to define objects and conditions. This can be particularly useful for accident prevention, real-time physical threat detection, and supply chain control inspections.
Cogniac supports multiple deployment setups, including on-premises, gateway, cloud, and hybrid approaches. A variety of contracts focused on manufacturing and industrial applications have helped boost the company’s year-to-year revenue by 78%. With a new series of financing led by Autotech Ventures and a solid technology platform, the company is expected to do particularly well in the years to come.
The platform uses custom-generated AI models for scenarios, crafted through the combination of imagery and user feedback. The system benefits from deep convolutional neural networks, which allow its models to learn new characteristics and adapt behavior based on previous input.
This continuous evaluation of feedback means that Cogniac can monitor the level of confidence of any new predictions, constantly searching for configuration variations for an optimal architecture. In consequence, manual intervention by data scientists or AI experts is ostensibly reduced.
Cogniac can achieve an accuracy of over 90% prior to human corrections. In order to establish a consensus among imbalanced datasets, the system employs a process in which multiple people review uncertain data in order to spotlight inherent biases. The platform is particularly transparent, as it can provide the provenance of all data and annotations.
Cogniac software has two government contracts and is participating in trials for a larger deployment. The US Army uses Cogniac software to analyze battlefield drone data, and the Arizona county sheriff’s department also utilizes it to identify when people cross the US-Mexico border.
This is not without controversy. Research has suggested that the last major effort to increase technology at the border may have contributed to an increase in the number of migrant deaths — as people tried to find (more treacherous) alternative routes in order to avoid detection. Civil liberty and digital privacy concerns have also been raised because integrating computer vision technology could propel a state of perpetual surveillance.
According to co-founder Bill Kish, these contracts are just a small portion of the company’s initiatives, which focuses mostly on manufacturing and industrial applications.
Cogniac doesn’t perform facial recognition or collect facial databases. This would line up with other companies focusing on visual object detection only. For example, earlier this year, IBM announced it was exiting the facial recognition market while urging a national dialogue on its use by law enforcement.
Manufacturing and Industrial Applications
Among Cogniac’s customers are some of the global leaders in the automotive, rail, logistics, packaging, and security industries. Two examples of the platform capabilities are mills and automobiles.
The average mill receives hundreds of truckloads of hardwood per day. Although these companies have nearly perfected the process of creating plywood, lumber, and OBS, there’s one area that has constantly shifting variables: the volume and variety of the logs received. Instead of manually inspecting them, a time-consuming task that is subjected to human error, Cogniac’s AI model has automatically tuned thousands of parameters to improve accuracy, lowering costs and providing higher yields.
Another successful case is automotive manufacturing. When miles of sheet metal are rolled out and deformed into body panels and stamping, the material can be prone to splitting. A manufacturing floor can stamp a car panel every four seconds, much faster than a person can inspect them. Also, training inspectors is costly and undetected splits could cause a whole vehicle to be lost. Deep Neural Networks like the one used by Cogniac learn what defects look like across different parts and can automatically hyper-optimize the parameters, making the task faster, cheaper, and more precise.
Cogniac built the first-of-a-kind AI that can perform a broad range of automated visual inspection tasks that were previously reserved for humans. As such, the company is uniquely positioned to transform many industries that depend on visual perception and semantic cognition.
The series B round announced this week is led by Autotech Ventures, with participation from Vanedge Capital, Yellowstone Ventures, Wing Venture Capital, and the George Kaiser Family Foundation/Energy Innovation Capital. The $10 million follows a $10.1 million series A from May 2019, bringing Cogniac’s total raised to over $20 million.
Since its series A, Cogniac has experienced 78% year-over-year (YoY) revenue and expanded its customer base to industry leaders. This new infusion of funding will allow the company to continue growing, improving visual inspection, and reinforcing its mission to provide exceptional solutions for global and North American customers.
About the Author
Yisela Alvarez Trentini is an Anthropologist + User Experience / Human-Computer Interaction Designer with an interest in emerging technologies, social robotics, and VR/AR.