What is Clear Lens?
Research/Service Description:
Apply self-supervised attention-based deep residual network to enhance the accuracy of camera anomaly detection and a small bottleneck ResNet structure to simplify the model for edge computing device implementation.
Research/Service Description:
Apply self-supervised attention-based deep residual network to enhance the accuracy of camera anomaly detection and a small bottleneck ResNet structure to simplify the model for edge computing device implementation.
Research/Service Outcome:
A real-time anomaly detection surveillance system with an accuracy of over 90 % in detecting four common anomalies.
 |
 |
 |
| AI ResNet structure data-draven methodlogy | System demo shows a normal class of picture outside a window | Four common anomaly classes: Haze, defocus, dirt and spray paint |
| Demo | Tools | Publication | Patent | Poster | Specification |