🧑‍🌾 Object-Detection Models: Weed Detection for Smart Agriculture

As part of my work with UntetherAI’s (now acquired by AMD) accelerator chip, I helped adapt object detection models for agricultural applications. One project involved a system mounted beneath tractors: as the tractor moves across the field, cameras capture real-time images of soil and plants. The AI model, running on Untether’s chip, detects weeds among the crops, and the system delivers a precise electrical zap to eliminate them without harming nearby plants.

My role was to restructure ONNX models so they could run fully on the chip, and to ensure that detection ran fast enough to keep up with the tractor in real-world field conditions. More details are in my blog post about autonomous vehicles, which was identical in terms of my process.