UMUHOZAALICE/Crop_classification

This project aims to classify different crop types in Rwanda such as banana, legumes, forests, etc. using deep learning model and transfer learning ## Crop_classification usinf Remote sensing Data # Abstact This study aims to address the challenge of crop classifiication in Rwanda using Rwanda Remote sensing data to classify different crop types such as banana, legumes, forests, etc. It uses using deep learning model combined with transfer learning using VGG 16 due to its effectiveness in image recognition tasks. the purpose of this study is to enhance the agricultural monitoring in form of farm management and resource allocation under accurate identification of these crops wth the se deep learning models. Thwe model were trained and tested with the promising result of about 95% accuracy. This finding highlights the potential of deep learning for precision agriculture in Rwanda. Providing accurate crop classification will potentially improve agriculture productivity with farm management optimization and resource allocation. This study contibutess to the integration of technology with traditional farming thereby enhancing accuracy and sustainablity through image based learning. The findings offer insights into agricultural productivity and sustainability, benefiting policymakers, MINAGRI, and researchers working towards agricultural advancements in Rwanda. RTI Rwanda crop-type dataset were used which consists of 6 classes for crops: forest, banana, other, structure, legumes and maize. # Exploratory Data Analysis This study adopts quantitave analysis method with remote sensing data to evaluate the effectiveness of deep earnign models in crop classification. The dataset, sourced from Research Triangle Institute (RTI) International, features remote sensing data captured via drones. It includes 5161 training images and 1290 validation images across six crop categories: banana, maize, legumes, structure, forest, and others. Each image in the training and validation datasets was recorded in a DataFrame with corresponding labels. Given the limited size of the dataset, data preprocessing and augmentation were essential. …