poojan-s/medicinal-leaf-classification-dataset

🌿 Medicinal Leaf Classification Dataset

An image dataset of 3 medicinal plant leaves — Aloe Vera, Neem, and Tulsi — used to train and evaluate deep learning classifiers.

Dataset Summary

Property	Value
Total Images	~7,380 (train + val)
Classes	3
Image Format	JPEG / PNG
Task	Image Classification

Classes

Index	Class	Train+Val Images	Test Images
0	Aloe Vera	—	183
1	Neem	—	453
2	Tulsi	—	284

Splits

Split	Images
Train	5,904
Validation	1,476
Total	7,380

A separate held-out test set of 920 images was used for final evaluation (not included here).

Folder Structure

train/
├── Aloe Vera/
├── Neem/
└── Tulsi/
val/
├── Aloe Vera/
├── Neem/
└── Tulsi/

Usage

from datasets import load_dataset

ds = load_dataset("poojan-s/medicinal-leaf-classification-dataset")

# Or load with torchvision / keras directly from folder structure

With Keras:

import keras

train_ds = keras.utils.image_dataset_from_directory(
    "train/",
    labels="inferred",
    label_mode="categorical",
    image_size=(224, 224),
    batch_size=32,
)

# Class order (alphabetical): ['Aloe Vera', 'Neem', 'Tulsi']
print(train_ds.class_names)

Associated Model

A MobileNetV3Large model fine-tuned on this dataset achieves 99% accuracy on the test set.

👉 poojan-s/mobilenetv3-medicinal-laef-classification

License

MIT