Detect fires early by identifying smoke or flames within the camera's field of view.
AI Wildfire Detection for Rapid Early Response
99%
Recognition accuracy
Achieves a high recognition accuracy of 99%
successfully identifying wildfire events in 73,442 out of 74,146 wildfire images collected from real cases.
0.0011
FPR (False Positive Rate)
Low false positive rate
Minimizes the rate of detecting non-wildfire events as wildfires, reducing unnecessary deployments and resource usage for non-critical situations.
#1 in Korea: Core Technologies for Anomaly Analysis
Wildfire Detection
Detects wildfires using outdoor images from RGB/IR cameras.During the daytime, the system focuses on detecting smoke in RGB images, while at night, it relies on IR images to identify flames as the primary feature.
Continual learning system
Continuously learns from error data occurring in CCTV environments.When deployed in real-world settings, the AI model is exposed to new data, including error cases such as false negatives and false positives. ALCHERA’s Anomaly Analysis Lab researches Active Learning, which automatically identifies errors in the data, and Continual Learning, which effectively incorporates these errors into AI’s learning process to improve overall model performance.
.svg)
.avif)
.svg)
.svg)
.svg)
.svg)
