Agriculture AI on FPGA 🌾

Overview

A multi-model AI system for smart agricultural decision making, deployed on Xilinx ZYNQ-7000 FPGA using hls4ml.

Models

Model	Task	Dataset	Accuracy
MLP	Crop Price Prediction	Kalimati Market	99.98%
Conv1D	Agriculture QA	Agriculture QA Dataset	TBD

Hardware

• Target Board: Xilinx ZYNQ-7000
• Tool: hls4ml
• Precision: ap_fixed<8,4>
• Reuse Factor: 1

Resources (ZYNQ-7000)

Resource	MLP	Available
LUTs	865	53,200
FFs	432	106,400
DSPs	8	220
BRAM	0	140

Dataset

• Kalimati Tarkari Market (Nepal) 2013-2021
• 197,162 records
• 132 unique crops

Hugging Face Models

• MLP: https://huggingface.co/MadhawaDG/farmer-price-predictor
• LSTM: https://huggingface.co/MadhawaDG/agriculture-lstm-zynq

Project Structure

agriculture-ai-fpga/ ├── mlp/ │ ├── farmer_price_model.keras │ ├── scaler.pkl │ ├── label_encoder.pkl │ └── hls_project/ ├── conv1d/ │ ├── conv1d_model.keras │ └── hls_project/ ├── notebooks/ │ ├── mlp_training.ipynb │ └── conv1d_training.ipynb └── README.md