I built a hybrid Machine Learning system that processes field video frame-by-frame to automate quality control (e.g., in sugarcane) and fruit counting, integrating these visual detections as features for tabular ensembles that predict yield and quality. In the live demo, I will show the full cloud inference workflow: from processing the raw video to classify quality (intact vs. damaged), the layered model architecture, and how we structure the data pipeline to bridge the gap between visual detection and final prediction.
