Enhancing real time tropical cyclone intensity estimation using YOLO-NAS algorithm with CLEO optimizer

Tropical cyclone intensity estimation is vital for disaster preparedness and mitigation. Owing to the processing constraints and intricate atmospheric patterns, traditional approaches frequently struggle with real-time accuracy. This study proposed an advanced model utilizing the YOLO-NAS method optimized by CLEO algorithm to enhance real-time estimation of cyclone intensities from satellite images. YOLO-NAS, which is distinguished by its superior object detection capabilities, is complemented by the CLEO optimizer that fine-tunes model parameters to improve detection accuracy while maintaining inference speed. The proposed method was evaluated on real-time satellite imagery, showing considerable improvement over traditional deep-learning models. Experimental results indicate that the presented approach attains higher precision, recall, and mean average precision (mAP) scores while also sustaining almost similar inference time and is thereby applicable for fast estimation of cyclone intensity. The proposed model demonstrates an improvement of 3.4, 4.44, 4.7, and 3.5%, in the accuracy, precision, recall, and mAP50 metrics, when compared with the unoptimized algorithm. The results provide evidence for the ability of YOLO-NAS, together with CLEO optimization, to provide an efficient and reliable tool for cyclone intensity monitoring in operational meteorology.