The recent advancements in machine vision systems such as 3D camera have proven to be a significant advantage in capturing critical real-time and processing over-the-edge computing devices that can give meaningful insights in real-time. In this paper we propose a unique setup to capture and process data real time using 3D machine vision camera system and applying deep learning models to detect the ore size, texture and characteristics using continuous rock flow video data. The insights are then instantly passed on to mill personnel for real time adjustment on mill throughput adjustment, blending and mixing etc. which results into significant optimization for overall mine to mill value chain.