1D_CNN_FPGA-Deployment

1D CNN model built with Keras/TensorFlow for text classification, trained on IMDB sentiment data and Agriculture Q&A dataset, with HLS4ML conversion for Zynq 7000 FPGA deployment.

1D CNN Text Classification

A complete pipeline for training a 1D Convolutional Neural Network (CNN) using Keras and TensorFlow on Google Colab.

What this repo contains

• Data loading and preprocessing pipeline
• 1D CNN model architecture with Conv1D, MaxPooling, Embedding layers
• Training on IMDB sentiment dataset (84% accuracy)
• Training on Agriculture Q&A dataset (87% accuracy)
• Model saved as .h5 format
• HLS4ML conversion to HLS C++ for FPGA deployment
• Vitis HLS project files for Xilinx Zynq 7000

Models

• Hosted on Hugging Face: MadhawaDG/agriculture-1d-cnn

Tech stack

• Python 3.12
• TensorFlow 2.19
• Keras
• hls4ml
• Google Colab
• Xilinx Vitis HLS
• Zynq 7000 FPGA (xc7z020clg484-1)

Pipeline

Dataset → Tokenize → Pad → Train 1D CNN → Save .h5 → Convert HLS → FPGA