ResearchProject case study

Plant Disease Detection

PaudhRog

AI tool that detects plant diseases via leaf images and suggests treatments for healthier crops.

Role Co-authoredIndividual contribution documented below

Explore the case study View Code

Built with

AIPythonTensorFlow

01 / PROJECT CONTEXT

What this project brings together

PaudhRog pairs a transfer-learned leaf-image classifier with a Flask prediction route and treatment suggestions, with the research also addressing class imbalance.

02 / THE PROBLEM

Problem

Spotting a plant disease early from a leaf is hard without an expert.

03 / APPROACH

Implementation approach

Transfer learning on InceptionResNetV2, trained on the PlantVillage dataset, served through a Flask /predict route that returns the disease class with treatment suggestions. The research also documents oversampling and undersampling to address class imbalance.

04 / IMPLEMENTATION WALKTHROUGH

How the experience works

A step-by-step view of the product flow, using the features and components documented for this project.

STEP 01Provide a leaf imageThe workflow begins with an image of a plant leaf.
STEP 02Prepare the imageImage preprocessing prepares the input for the transfer-learning model.
STEP 03Run InceptionResNetV2The documented model uses InceptionResNetV2 transfer learning on the PlantVillage dataset.
STEP 04Serve the predictionA Flask /predict route serves the model output to the web application.
STEP 05Present the resultThe application returns a disease class with treatment suggestions.

05 / CONTRIBUTION

My contribution

Implemented the InceptionResNetV2 transfer-learning pipeline in TensorFlow, including the oversampling and undersampling strategy used to correct class imbalance across the PlantVillage set, and built the Flask serving layer and web interface around it. Co-authored the manuscript, drafting the methodology and results sections.

06 / SYSTEM DESIGN

Architecture at a glance

This diagram summarizes the documented system boundaries. The walkthrough above explains the user and processing flow in more detail.

Client
Web UIWeather widget
Server
Flask app/predict route
Model
PreprocessingInceptionResNetV2
Data
PlantVillage dataset

Leaf image in, disease class out. Transfer learning over InceptionResNetV2, served through Flask.

07 / DESIGN NOTES

Trade-offs & constraints

These are practical implications of the architecture described in the supplied project notes, not claims about a formal design review.

Transfer learning reuses the InceptionResNetV2 representation rather than training an image model from scratch. The model's depth and size still need to be considered when packaging and serving inference.

08 / CAPABILITIES

Key features

Image-based disease detection
Treatment recommendations
Multi-crop support

Opening case studyPreparing project details and architecture

How the experience works

A step-by-step view of the product flow, using the features and components documented for this project.

STEP 01Provide a leaf imageThe workflow begins with an image of a plant leaf.

STEP 02Prepare the imageImage preprocessing prepares the input for the transfer-learning model.

STEP 03Run InceptionResNetV2The documented model uses InceptionResNetV2 transfer learning on the PlantVillage dataset.

STEP 04Serve the predictionA Flask /predict route serves the model output to the web application.

STEP 05Present the resultThe application returns a disease class with treatment suggestions.