 This study proposes a pre-trained feature-aggregated convolutional neural network approach for detecting and monitoring overshooting tops, which plays a crucial role in climate change and severe weather conditions. The proposed model considers both physical and spatial characteristics of OTs using multi-channel data from Geo, CompSat, 2A advanced meteorological imager, GK2AAMI, over East Asia. Six schemes were evaluated, resulting in a probability of detection, POD, of 92.1%, a false alarm ratio, FAR, of 21.5%, and a critical success index, CSI, of 0.7%, which is significantly improved compared to the existing CNN-based OT detection model. This article was authored by Yoo-Hoon Lee, Mayee Kim, Jung-Ho Im, and others.