This study introduced an automated edge-computing system to evaluate sweet corn maturity, addressing a vital need for precise harvest timing to maximize crop value in the Philippines. Operating on a Raspberry Pi 4 with an external camera, the standalone device utilized a robust two-stage deep learning pipeline. It employed YOLOv7 for real-time object detection alongside a ResNet-50 convolutional neural network to accurately classify the corn as mature, immature, or over-mature. Trained on a custom dataset of 2,400 locally sourced images, the system demonstrated exceptional reliability in a controlled environment. It achieved a flawless 100% detection rate and an 88.89% multi class classification accuracy, surpassing the 85% target. Ultimately, this framework provided farmers a highly practical solution to standardize crop quality and optimize their harvest schedules.