bf369/BeejX-Crop-Disease-Dataset

BeejX Crop Disease Dataset

Edge-AI Data Solution for Crop Disease Recognition (Offline & Mobile-First).

Dataset Overview

The BeejX Crop Disease Dataset is an extensive, multi-crop image classification dataset curated for training lightweight Edge-AI models. The dataset is specifically structured and balanced to train models like MobileNetV2 (Quantized), enabling real-time, offline inference on low-end Android devices in rural agricultural areas.

This dataset contains a flattened and processed directory structure, ready to be plugged into standard image classification pipelines like tf.keras.utils.image_dataset_from_directory.

Crop & Disease Classes (50 Classes)

This dataset covers 50 distinct classes spanning healthy and diseased states across multiple vital crops:

• 🥭 Mango: Anthracnose, Die Black, Gall Midge, Powdery Mildew, Healthy
• 🥒 Cucumber: Bad, Good
• 🌾 Mandua (Finger Millet): Blast, Downy, Mottle, Rust, Seedling, Smut, Wilt, Healthy
• 🌿 Mulberry: ChiangMai60, RedKing, WhiteKing, BlackOodTurkey, TaiwanStraberry, BlackAustralia, Buriram60, Kamphaengsaeng42, TaiwanMeacho, ChiangMaiBuriram60
• 🫘 Pigeon Pea: Leaf Spot, Leaf Webber, Sterilic Mosaic, Healthy
• 🥔 Potato: Bacteria, Fungi, Nematode, Pest, Phytopthora, Virus, Healthy
• 🌹 Rose: Black Spot, Dry Leaf, Leaf Hole, Healthy Leaf
• 🎋 SugarCane: Mosaic, RedRot, Rust, Yellow, Healthy
• 🫚 Turmeric: Dry Leaf, Leaf Blotch, Rhizome Disease Root, Healthy Leaf & Root

Original Data Sources & Citations

This dataset is an aggregation of multiple open-source agricultural datasets, carefully cleaned and standardized for the BeejX pipeline. If you are researching specific crops, please refer to the original raw sources:

• Mango Dataset: Mendeley Data
• Sugarcane Dataset: Kaggle Dataset
• Potato Leaf Dataset: Mendeley Data
• Cucumber Leaf Dataset: Mendeley Data
• Rose Leaf Dataset: Mendeley Data
• Mulberry Leaf Dataset: Mendeley Data
• Lemon Leaf Dataset: Mendeley Data
• Sesame Leaf Dataset: Mendeley Data
• Arecanut Disease: Kaggle Search Reference
• Additional Mendeley References:
  • Source 1 (bd553pdtny)
  • Source 2 (45djgf3p96)
  • Source 3 (g46dvrcvwn)

Tech Stack & Data Processing Pipeline

As part of the BeejX modeling pipeline, this dataset is designed to be paired with:

• Custom Loaders: Handled natively by custom BeejXDataLoader with Real-time Augmentations (Rotation/Zoom).
• Imbalance Handling: Optimized for Algorithmic Class Weighting (via sklearn) during model training.
• Data Version Control (DVC): Originally tracked and structured using DVC for reproducibility.

Usage Request

If you use this aggregated dataset for academic or commercial agricultural AI applications, please credit the BeejX LeafModel project and ensure you comply with the licenses of the original datasets listed above.