BrandonFors/effnetv2_s_plant_disease

Model Card for Model ID

Classifies images of leaves as healthy or diseased (also specifies type of plant as well as disease if within dataset)

Model Details

effnetv2-s architecture

pretrained from: https://docs.pytorch.org/vision/stable/models/generated/torchvision.models.efficientnet_v2_s.html#torchvision.models.efficientnet_v2_s With .DEFAULT weights

Uses

Plant disease classifcation Finetune to a larger dataset/dataset with more classes?

How to Get Started with the Model

Download the modeling.py and configuration.py files from the repo

Run:

from modeling import EffNetPlantDiseaseClassification
test_model = EffNetPlantDiseaseClassification.from_pretrained("BrandonFors/effnetv2_s_plant_disease").to(device)

There's no pipeline (sorry), so the model must be run with a standard pytorch prediction loop

Training Details

Trained for 5 epochs with Adam(lr=1-e4)

Training Data

https://huggingface.co/datasets/BrandonFors/Plant-Diseases-PlantVillage-Dataset

Training Procedure

Refer to notebook in github repo

Preprocessing

Use the torch transforms provided by torchvision

# Get model weights from torchvision
import torchvision
### EffNetV2 - S
effnetv2_s_weights = torchvision.models.EfficientNet_V2_S_Weights.DEFAULT
effnetv2_s_auto_transforms = effnetv2_s_weights.transforms()

Training Hyperparameters

fp16 mixed precision

Evaluation

Refer to the TF Training Metrics tab

Testing Data, Factors & Metrics

Testing Data

https://huggingface.co/datasets/BrandonFors/Plant-Diseases-PlantVillage-Dataset

Metrics

Accuracy - Get an overall picture of model effectiveness F1 Score - Account for classes with a smaller image amount

Results

[More Information Needed]

Summary

Hardware

T4 GPU

Software

Google Colab