Lightweight AI Model Detects Forest Fires Faster With Fewer Parameters

Researchers at Shanxi Agricultural University have developed YOLOv8-CNR, a lightweight AI model for detecting forest fire flames and smoke that targets real-time monitoring. Built by modifying the YOLOv8n object detector at three points—backbone feature extraction, the neck’s multi-scale feature fusion, and the loss function—the model reached 79.8% mAP50 and 39.4% mAP50-95 on a synthetic forest wildfire benchmark. It uses 2.52 million parameters and processes 238.1 frames per second. The team says flames and smoke are visually difficult in different ways, while forest backgrounds add shadows and textures that can mimic targets.