Deep learning has proved to be a runaway success in the area of computer vision. Pioneering work by Yann LeCun and others on convolutional neural network architectures, spurred by benchmark challenges such as ImageNet, have been pivotal in pushing the state-of-the-art towards human-level or better performance in key tasks such as object detection and classification.
Although such network architectures are almost by day multiplying and pushing the envelope, enterprise applications have been limited to areas close to the technology sector’s heart like self-driving cars. This belies untapped opportunity to utilize these methods in monitoring tasks where previously only a battery of dedicated human observers could possibly be effective.
In the video, we monitor traffic in a section of our workspace with a camera. Detections of objects are shown in bounding boxes, including detections of persons. These are produced frame by frame on live video through a performant ensemble of algorithms.
By pre-defining areas of interest on such a video feed, understanding the flow of human traffic across them or how busy each area gets and when is now as simple as logging detections by the algorithm. Applied at scale, the technology can unlock value from investments such as surveillance systems hitherto unimagined.
This is a simple example of a computer vision task on data from hardware infrastructure already likely to exist. Limited only by the latter, it becomes possible to algorithmically build on this initial detection to track an individual’s journey over a span of time and across multiple video feeds or to monitor posture and even emotion or satisfaction level to gain a deep understanding of behavior in real-time.
Anomalous behaviors in cases like customer dissatisfaction can be flagged for human review and possible intervention, saving employees’ time, while behavior patterns conducive to desirable outcomes can be identified ahead of time to inform predictive models for metrics capturing revenue, brand loyalty, and other performance indicators.Trabeya’s intelligent vision solution enables behavioral insights in multiple industry settings and use cases. As with any artificial intelligence solution, the strong baseline achieved on generic tasks through transferability can often be just the springboard for further domain-specific training. We welcome you to reach out for further information on how our solution can add value to your problem domain.
Deep learning has proved to be a runaway success in the area of computer vision. Pioneering work by Yann LeCun and others on convolutional neural network architectures, spurred by benchmark challenges such as ImageNet, have been pivotal in pushing the state-of-the-art towards human-level or better performance in key tasks such as object detection and classification.
Although such network architectures are almost by day multiplying and pushing the envelope, enterprise applications have been limited to areas close to the technology sector’s heart like self-driving cars. This belies untapped opportunity to utilize these methods in monitoring tasks where previously only a battery of dedicated human observers could possibly be effective.
In the video, we monitor traffic in a section of our workspace with a camera. Detections of objects are shown in bounding boxes, including detections of persons. These are produced frame by frame on live video through a performant ensemble of algorithms.
By pre-defining areas of interest on such a video feed, understanding the flow of human traffic across them or how busy each area gets and when is now as simple as logging detections by the algorithm. Applied at scale, the technology can unlock value from investments such as surveillance systems hitherto unimagined.
This is a simple example of a computer vision task on data from hardware infrastructure already likely to exist. Limited only by the latter, it becomes possible to algorithmically build on this initial detection to track an individual’s journey over a span of time and across multiple video feeds or to monitor posture and even emotion or satisfaction level to gain a deep understanding of behavior in real-time.
Anomalous behaviors in cases like customer dissatisfaction can be flagged for human review and possible intervention, saving employees’ time, while behavior patterns conducive to desirable outcomes can be identified ahead of time to inform predictive models for metrics capturing revenue, brand loyalty, and other performance indicators.Trabeya’s intelligent vision solution enables behavioral insights in multiple industry settings and use cases. As with any artificial intelligence solution, the strong baseline achieved on generic tasks through transferability can often be just the springboard for further domain-specific training. We welcome you to reach out for further information on how our solution can add value to your problem domain.