Multi-Crop Plant Disease Classification with EfficientNet-B4 & AlexNet

Introduction

Plant diseases threaten food supply everywhere, cutting yields, raising costs, and pushing farmers toward heavy, reactive spraying. Checking crops by hand is slow and easy to get wrong.

Image software can recognise leaf diseases across several crops from ordinary photos, helping catch problems like apple scab, grape black rot, and corn rust before they spread. This project builds such a system and wraps the best model in a simple web tool where a user can upload a leaf photo and get an instant result.

Problem Statement

Diagnosing plant disease by eye has real drawbacks:

1. It is subjective: experts can disagree on early symptoms.
2. It is slow and hard to scale across large farms and orchards.
3. Different crops need different expertise.
4. Many rural areas simply do not have enough specialists.

An automated helper also has to cope with look-alike symptoms and still run on affordable hardware.

Objectives

1. Recognise leaf diseases across several crops from photos, in one model.
2. Compare a few image models to see which is most reliable.
3. Use a balanced: well-prepared image set for fair training.
4. Judge the models on clear: side-by-side measures.
5. Wrap the best model in a simple web tool for hands-on leaf diagnosis.

Dataset

The work is built around a balanced set of leaf images spanning several crops and diseases. Depending on the client, this can be a public plant-disease benchmark, licensed or private field photos, images from other trusted sources, or a custom set gathered for the project. Images are cleaned, balanced across classes, resized, and expanded with standard variations that mimic real capture conditions.

Methodology

The project trains and compares four image models on the same prepared data: Basic CNN, Modified AlexNet, EfficientNet-B0, and EfficientNet-B4, from a simple baseline through to efficient, modern designs that reuse knowledge from earlier training.

The novelty is closing the loop from experiment to something usable. Rather than stopping at a benchmark, the project weighs accuracy against efficiency across several crops at once and turns the winner into a hands-on tool where anyone can upload a leaf photo and get an instant diagnosis, a practical differentiator over benchmark-only studies.