Undergraduate research
Precision Agriculture Computer-Vision Research
Computer-vision research using custom image datasets and YOLOv8 to detect agricultural disease.
- Role
- Undergraduate Research Fellow
- Organization
- Morehead State University
- Timeline
- September 2023–May 2024
- Status
- Completed
- ~99.5%
- Reported model accuracy; approximate résumé figure
Problem
Agricultural disease detection requires models that can distinguish visually similar conditions across uneven, real-world image data.
Context
- The undergraduate research focused on custom datasets and predictive models for agricultural disease detection.
- The work covered data preparation, model training, classification, evaluation, and predictive analysis.
Requirements
- Create useful training data from a custom image collection.
- Evaluate model behavior rather than relying on a single training run.
- Communicate technical findings to an academic audience.
My role
- Prepare and classify image data using Roboflow and Python workflows.
- Train and evaluate YOLOv8 computer-vision models.
- Use Pandas and NumPy to prepare data, analyze results, and review errors.
- Document predictive results and sources of error in the research process.
Process
- 01
Curate
Built and organized a custom agricultural image dataset.
- 02
Prepare
Applied preprocessing and classification to create model-ready data.
- 03
Train
Used YOLOv8 and Roboflow within an iterative training workflow.
- 04
Evaluate
Reviewed model performance and error patterns before reporting results.
Tools
Challenges
- 01
Creating consistent labels across visually similar disease classes.
- 02
Separating a strong reported result from broader claims about field deployment.
- 03
Explaining model behavior in a way that connects technical work to agricultural use.
Outcome
The research produced a workflow for preparing images, training models, and reviewing disease-detection results.