Artificial Intelligence
Artificial intelligence is a wide and deep field that encompasses many underlying technologies and methods, including machine learning and deep learning. According to Andrew Moore, Dean of Computer Science at Carnegie Mellon University, in an interview with Forbes in 2017, “Artificial intelligence is the science and engineering of making computers behave in ways that, until recently, we thought required human intelligence.” The evolution of AI will transform industries and the human experience. Being able to harness this power, both technically and ethically, positions organizations for immense growth and the realization of their innovative visions. Dignari has broad experience in applied AI and mission driven research and development.
Computer Vision
We create solutions utilizing extensive computer vision capabilities, allowing a user to leverage information gained from cameras interpreting their field of view. This includes body worn technologies, existing surveillance cameras, or mobile phones.
Customer Requirements
Reduce the occurrence of occlusions when performing face recognition Minimize the need for human operators Maintain optimal system throughput Dignari Solution: Dignari developed a computer vision prototype utilizing open source software, layered with a custom developed machine learning model, to detect and alert on the presence of occlusions. We trained an artificial neural network object detection model to be able to detect client-defined occlusions and to alert an individual in the camera’s field of view to remove the detected item. Additionally, the custom model was deployed to the camera directly to further test edge computing capabilities and the ability to use the camera’s platform for ML functionality.
Outcome for the Customer
The customer gained insight into the impact AI/ML could make in their mission environment and how it might augment their current operations. The development of these capabilities allows them to implement intelligence at the edge and to have computers/cameras engage with their end users without the need for a human operator in the loop.
Natural Language Processing - We create solutions that allow system control based on voice commands. This includes voice recognition, speech-to-text, translation, speaker identification, and speaker diarization.
Custom Machine Learning Model Training - We customize ML models based on mission need and use case applicability, using techniques such as Tensorflow transfer learning. We also analyze the mission fit of deployment of custom models to edge computing environments.