How chinmoydeb/mobilenetv2-plant-disease-detection works
This repository presents a PlantDoc plant-disease classifier built by adapting a pretrained MobileNetV2 model. The documented workflow uses PlantVillage pretraining, then trains a custom head with the backbone frozen before progressively fine-tuning deeper blocks on PlantDoc, followed by evaluation. The README also describes experiment tracking and an inference example. The notebook is the only listed implementation source, but its contents were not sampled, so internal call relationships cannot be verified; edges below reflect README-documented workflow and capabilities.
Main entry points: plantdoc_transfer_learning.ipynb Frozen-head training · plantdoc_mobilenetv2.pth Saved model
Components 14
Data and Adaptation
- PlantVillage data
- PlantDoc data
Model Training
- Pretrained MobileNetV2
- Frozen-head training
plantdoc_transfer_learning.ipynb - Progressive fine-tuning
plantdoc_transfer_learning.ipynb - Classification head
plantdoc_transfer_learning.ipynb - Saved model
plantdoc_mobilenetv2.pth
Evaluation and Use
- PlantDoc evaluation
plantdoc_transfer_learning.ipynb - Experiment tracking
plantdoc_transfer_learning.ipynb - Image preprocessing
plantdoc_transfer_learning.ipynb - Disease prediction
plantdoc_transfer_learning.ipynb
Other components
- Training user
- Weights & Biases
- Image user
Connections 15
- PlantVillage data → Pretrained MobileNetV2: pretrains
README.md - Pretrained MobileNetV2 → Frozen-head training: initializes
README.md - Frozen-head training → Classification head: trains head
README.md - Frozen-head training → Progressive fine-tuning: continues as
README.md - PlantDoc data → Frozen-head training: trains on
README.md - PlantDoc data → Progressive fine-tuning: fine-tunes on
README.md - Progressive fine-tuning → Saved model: saves
- Progressive fine-tuning → PlantDoc evaluation: evaluates
README.md - Training user → Frozen-head training: runs training
README.md - Training user → Experiment tracking: tracks runs
README.md - Experiment tracking → Weights & Biases: logs metrics
README.md - Image user → Image preprocessing: provides image
README.md - Image preprocessing → Disease prediction: prepares input
README.md - Saved model → Disease prediction: loads model
- Disease prediction → Image user: returns prediction
README.md