AgriVision MobileNetV2 Edge TPU Model Card

Model Overview

AgriVision MobileNetV2 Edge TPU is a lightweight plant-health image classification model prepared for smart agriculture demonstrations, edge-AI research, and reproducible single-board-computer deployment workflows.

The model card documents a field-facing inference pattern in which crop or leaf images are processed locally on embedded hardware. It is intended to support transparent research demonstrations, classroom exhibition workflows, and prototype agricultural monitoring systems.

Intended Use

• Plant-health image inference for smart agriculture research prototypes.
• Edge deployment demonstrations using Raspberry Pi 4 and Coral USB Edge TPU.
• Model-release documentation for reproducible training and deployment workflows.
• Educational demonstrations of sensor-controlled irrigation and local AI inference.

Inputs And Outputs

The expected input is an image prepared through the AgriVision preprocessing workflow. The expected output is a plant-health class prediction with confidence information suitable for research dashboards, logs, or demonstration interfaces.

Deployment Context

The deployment target is an edge-AI stack based on Raspberry Pi 4, Google Coral USB Edge TPU, TensorFlow Lite, and local sensor workflows. The design emphasizes low-latency local inference and reduced dependency on cloud connectivity.

Limitations

• Model outputs are research and demonstration signals, not agronomic advice.
• Performance can vary with crop type, lighting, camera quality, and image quality.
• Field use requires independent validation before operational decisions are made.
• The model should not be used as the sole basis for pesticide, irrigation, or crop-treatment decisions.

Access

The public model repository is hosted on Hugging Face under the PeachBot AI organization and is listed with the Apache 2.0 license.