We don’t know when accidents happen, such as a person falling in an unnoticeable location, property damage at night, or violence in the area. Security guards’ monitoring capacity is limited, leading to failed cases of accident prevention.
The implementation of Asilla’s product can strengthen security 24/7. Replacing security guards, a security camera continuously monitors the facility and alerts security guards in case of abnormal activities for immediate actions.
Asilla has developed a proprietary algorithm for tracking people in time-series images. Improvement of tracking performance allows for the examination of individual detection ID, even when the target is temporarily hidden.
Thanks to an autonomous learningfunction within, each edge AI device can generate an AI model based on its independent experience in the environment. As a result, each device can act as a substitute to a section of the cerebral cortex (or the fusiform gyrus area), allowing for highly precise identification of body parts. As it learns and performs autonomously, the AI can improve its accuracy without human assistance.
Open-source systems such as OpenPose have high computation cost, leading to significant implementation cost when real-time performance is required. Asilla has succeeded in transferring information from a large high-precision model to a compact one, resulting in considerable weight reduction. Thanks to this method, one server can process around 80 cameras simultaneously.
Through self-learning, AI models with optimized views for each angle are automatically generated. Detection of “abnormal behaviors” (deviations from normal behaviors) can prevent accidents.
|“Behavior identification system” – Patent number: 6793383||Technology to track the same person with multiple cameras|
|"Behavior inference device" - Patent number: 6525181||Algorithm to detect backgrounds and adjust likelihood of behaviors|
|"Behavior inference device" - Patent number: 6525179||Algorithm to recognize behaviors from joint coordinate values|
|"Target number identification device" - Patent number: 6525180||Algorithm to track people in an area in chronological order|
|"Suspicious/abnormal subject detection device" - Patent number: 6647489||Detection of abnormal behaviors of people in an area|
|"Abnormal behavior detection device" - Patent number: 6692086||Detection of abnormal behaviors from behaviors of surrounding people|
|"Suspicious/abnormal subject detection device" - Publication number: 2020-091856||Detection of abnormal behaviors from behaviors of surrounding people|
For product inquiries and consultations