Academic / computer vision · 2025-08-17
PlantDoc: Plant Disease Classification
Developed a plant-disease classifier using an attention-augmented ResNet18 and inspected model behavior with evaluation visualizations.
- Role
- ML project developer
- Contribution
- Developed a plant-disease classifier using an attention-augmented ResNet18 and inspected model behavior with evaluation visualizations.
- Evidence
- View public source ↗
Problem
Plant-disease classification is a useful setting for studying visual recognition, class confusion, and model interpretability. This academic project explores classification from labeled plant images.
Approach
I developed a ResNet18-based classifier with convolutional block attention modules (CBAM), and used PyTorch for training and evaluation. The project includes visual inspection of predictions and attention behavior.
Evaluation
The public repository contains the implementation and evaluation artifacts. Confusion matrices and GradCAM examples help examine which classes the model confuses and which image regions influence predictions.
Outcome
The project produced a classifier and an inspectable set of evaluation artifacts. It demonstrates model development and analysis on the project dataset.
Limitations
Dataset performance does not establish field accuracy on unfamiliar plants, lighting conditions, or cameras. This is an academic model, not a validated agricultural diagnostic product. Review the repository and its evaluation setup before interpreting individual metrics.