Green-Machine: FPGA Management System
← back to projectsOverview
Green-Machine is a smart, domestic mini-greenhouse management system built for the Intel InnovateFPGA design contest (2021–2022), together with Pahan Mendis and Chathuni Wijegunawardana.
Running on an Intel Cyclone V SoC (Terasic DE10-Nano), the system automates day-to-day plant care: irrigation driven by soil-moisture thresholds, grow-light control, and temperature and humidity regulation through ventilation. Sensor readings and plant status stream to the cloud (Microsoft Azure IoT Hub) for remote monitoring, and the whole system runs under 18 W. In testing it saved 20–40% of water compared to manual watering while growing fully organic produce.
My Contribution: On-FPGA Plant Health Detection
I built the machine learning part of the system: a convolutional neural network that watches the plants through a camera and detects abnormalities such as disease or wilting.
- Designed a compact 5-layer CNN (three convolution + max-pooling stages followed by two fully connected layers) suited to the FPGA’s resource constraints.
- Implemented the network for the Cyclone V SoC using the OpenCL SDK for Intel FPGAs, so inference runs on-device rather than in the cloud.
- Trained and validated the model on a dataset of around 3,500 plant images, reaching roughly 80% accuracy in abnormality detection.
Recognition
Green-Machine won the Gold Award at the Asia-Pacific regional final and the Global Silver Award at the global grand final of the Intel InnovateFPGA design contest.
Full build details, including the hardware design and sensor suite, are on the InnovateFPGA team page.