Radishes, a fast-growing root vegetable rich in vitamins and minerals, are a lifeline for farmers in Bangladesh and other regions. However, these crops face relentless threats from leaf diseases that can devastate yields and livelihoods.
Traditional methods of identifying diseases—such as visual inspections by farmers—are often inaccurate, leading to wasted resources and failed harvests.
To address this, researchers at East West University in Bangladesh have developed a groundbreaking dataset of 2,801 high-resolution images capturing healthy radish leaves and four major diseases. Published in 2025, this resource aims to train AI tools for precise, real-time disease detection.
Why Combating Radish Leaf Diseases is Critical for Farmers
Radishes are not just a dietary staple—they are an economic pillar for millions of smallholder farmers. These vegetables grow rapidly, often ready for harvest in under a month, making them ideal for regions with short growing seasons.
However, diseases like Black Leaf Spot (caused by bacteria) and Mosaic Virus (spread by insects) can slash yields by 40–60% if untreated.
For context, Black Leaf Spot refers to dark, water-soaked lesions caused by the pathogen Xanthomonas campestris, which attacks leaf tissues and reduces photosynthesis.
Farmers relying on traditional diagnosis methods frequently misidentify these diseases. For example, Downy Mildew—a fungal infection marked by yellow patches and white fungal growth—is often confused with nutrient deficiencies.
Such errors lead to unnecessary pesticide use, costing farmers 15–20% of their annual income and harming ecosystems. This cycle underscores the urgent need for accurate, accessible diagnostic tools.
Inside the Dataset: A Closer Look at the 2,801 Images
The dataset, hosted on Mendeley Data, is a meticulously curated collection of 2,801 images divided into five categories:
- Healthy Leaves (613 images): Vital for comparing diseased samples.
- Black Leaf Spot (526 images): Caused by Xanthomonas campestris, this disease creates dark lesions.
- Downy Mildew (601 images): Triggered by the fungus Peronospora parasitica, it thrives in humidity.
- Mosaic Virus (548 images): A viral infection spread by aphids (tiny insects).
- Flea Beetle Damage (513 images): Results from beetles chewing tiny holes in leaves.
Each image was captured using a Realme 9 5G smartphone with a 48MP camera, ensuring sharp, detailed photos. To maintain consistency, leaves were photographed in a lab-like setup with controlled lighting and a black backdrop.
Researchers took 15–20 shots per leaf from six angles, capturing nuances like texture and color variations. This approach ensures the dataset reflects real-world conditions, making it invaluable for training reliable AI models.
Breaking Down the Diseases: Causes, Symptoms, and Impact
Understanding these diseases is key to appreciating the dataset’s value. Black Leaf Spot, caused by the bacteria Xanthomonas campestris, appears as dark, water-soaked spots with yellow edges. These lesions spread rapidly, weakening plants and reducing root size by 30–40%.
In contrast, Downy Mildew—a fungal disease—creates yellow patches on leaf tops and fuzzy white growth underneath.
It flourishes in humid climates and can destroy 50–70% of crops if untreated.
Mosaic Virus, transmitted by aphids, causes mottled yellow-green patterns and curled leaves. Infected plants become stunted and vulnerable to secondary infections.
Lastly, Flea Beetle Damage—caused by tiny beetles chewing holes—can kill seedlings and reduce the market value of mature plants by 20–30%. Early detection of these diseases is critical, as delays allow pathogens to spread unchecked.
Building the Dataset: A Collaborative Effort
Creating this resource required collaboration between researchers, farmers, and agronomists. The team visited 12 radish farms in Brahmanbaria, Bangladesh—a region chosen for its high disease prevalence. Local farmers helped identify diseased plants, while agronomists confirmed diagnoses using lab tests.
In the lab, leaves were photographed under controlled lighting to avoid glare. A black velvet backdrop helped isolate each leaf, highlighting details like lesions or holes.
Images were then resized to 700×700 pixels and organized into folders labeled by disease type. Quality checks removed blurry or overexposed photos, ensuring only the clearest images were included.
Challenges in Data Collection and Limitations
Despite its rigor, the project faced hurdles. Locating enough diseased plants took three months, requiring close coordination with farmers. Early attempts to photograph leaves in direct sunlight failed due to glare, forcing a shift to indoor setups.
The dataset also has limitations. It focuses on diseases common in Bangladesh, such as Black Leaf Spot, but excludes others like Alternaria Blight (a fungal disease prevalent in other regions).
Additionally, the team did not use data augmentation—a technique where existing images are altered (e.g., rotated or flipped) to create new training examples. Augmentation can improve AI model accuracy, but its absence here means future researchers might need to expand the dataset.
Transforming Farming with AI: Applications and Benefits
This dataset opens doors to transformative tools. For instance, convolutional neural networks (CNNs)—a type of AI model that analyzes visual data—could be trained to detect diseases from smartphone photos.
Models like MobileNet (used in rice disease detection) or ResNet (known for high accuracy) might achieve similar success here.
- Farmers could use AI-powered apps to scan leaves and receive instant diagnoses, similar to Plantix, an app that reduced pesticide use by 25% in Indian farms.
Early warnings could also help farmers apply targeted treatments, saving costs and minimizing environmental harm. Over time, integrating weather data might enable apps to predict outbreaks, offering proactive solutions.
Ethical and Practical Steps for Widespread Adoption
The dataset’s CC BY 4.0 license—a Creative Commons agreement allowing free use with proper credit—encourages global collaboration. However, success hinges on educating farmers about AI tools. Workshops or community programs could teach them to use apps effectively, ensuring the technology reaches those who need it most.
Environmentally, reducing pesticide misuse aligns with the UN’s Zero Hunger initiative, which promotes sustainable farming. By curbing chemical runoff, AI tools could protect soil and water quality, benefiting ecosystems and human health.
The Future of Farming: AI-Driven Solutions and Global Collaboration
This dataset is a stepping stone toward precision agriculture—a farming approach that uses technology to optimize crop health.
Future research could expand the dataset to include diseases from India, China, or Brazil, improving AI models’ global relevance. Partnerships with tech companies might yield affordable apps tailored to small farmers, while governments could fund digital infrastructure.
As one researcher noted, “Democratizing AI for farmers isn’t just innovation—it’s justice.” By bridging the gap between technology and tradition, this work offers hope for healthier crops, stronger economies, and food security in a changing climate.
Conclusion
Radish farming, like all agriculture, faces mounting challenges from climate change and disease. Traditional methods alone are insufficient, but tools like this dataset show how technology can fill gaps. By turning smartphones into diagnostic devices, farmers gain power against invisible threats.
While hurdles like global accessibility remain, this innovation marks a leap toward equitable, sustainable farming. For families depending on radish crops, it’s more than progress—it’s a lifeline.
Power Terms
Dataset Collection: The process of gathering structured information, such as images, for research. In this study, 2,801 high-quality images of radish leaves (healthy and diseased) were collected using a smartphone. This step is crucial because accurate data helps train AI models to detect diseases, which benefits farmers and agriculture.
Image Analysis: Examining images to identify patterns or features. Here, photos of radish leaves were analyzed to spot signs of diseases like spots or discoloration. This helps automate disease detection, replacing time-consuming manual inspections by farmers.
Deep Learning Models: Advanced computer programs that learn from data. These models, like the ones trained in this study, can analyze leaf images to classify diseases automatically. For example, they might detect a fungal infection by recognizing specific visual patterns.
Black Leaf Spot: A bacterial disease caused by Xanthomonas campestris. It creates dark, water-soaked spots with yellow rings on radish leaves. Early detection using images can prevent crop loss, as farmers can apply targeted treatments.
Downy Mildew: A fungal disease caused by Peronospora parasitica. Infected leaves develop yellow-brown patches on top and white-gray fuzzy growth underneath. This disease reduces radish yield, so identifying it early helps farmers save crops.
Mosaic Virus: A viral infection that causes yellow-green mosaic patterns and curling on leaves. It spreads quickly, so detecting it early (via image analysis) allows farmers to remove infected plants and protect the rest.
Flea Beetle: A tiny pest that chews small holes in radish leaves. Severe infestations can kill young plants. Image datasets help train models to recognize these holes, alerting farmers to act before the crop is ruined.
Convolutional Neural Network (CNN): A type of AI model that processes images. It works like layers of filters, first detecting edges or spots, then combining them to recognize complex patterns (e.g., disease symptoms). CNNs are widely used in plant disease studies.
MobileNet: A lightweight CNN model designed for mobile devices. In a similar study, MobileNet achieved 90% accuracy in detecting rice diseases. Its efficiency makes it suitable for real-time farming apps.
Pathogen: A microbe (bacteria, fungus, or virus) that causes disease. For example, Xanthomonas campestris is the pathogen behind Black Leaf Spot. Identifying pathogens helps farmers choose the right pesticide.
Xanthomonas campestris: A bacterium that attacks radish leaves, causing Black Leaf Spot. It spreads through water or insects, making early detection vital to stop outbreaks.
Peronospora parasitica: A fungus responsible for Downy Mildew. It thrives in humid conditions and damages leaves, reducing photosynthesis. Detecting its fuzzy growth early can save entire crops.
Chlorosis: Yellowing of leaves due to nutrient loss or disease. In radishes, chlorosis is a symptom of Mosaic Virus. Spotting this color change in images helps diagnose the virus quickly.
Lesions: Damaged, discolored areas on leaves caused by disease. For example, Black Leaf Spot creates dark lesions. AI models analyze these marks to classify the disease type.
Mycelium: The thread-like part of a fungus (e.g., Peronospora parasitica). It appears as white-gray fuzz under leaves during Downy Mildew. Detecting mycelium in images confirms fungal infection.
Data Augmentation: Techniques to expand a dataset, like flipping or rotating images. The researchers did not use this but noted that doing so might improve model accuracy in future work.
Real-time Disease Detection: Instant diagnosis using AI models, often via mobile apps. While not implemented here, such tools could let farmers photograph leaves and receive immediate advice.
Ethical Guidelines: Rules ensuring research is conducted responsibly. This study followed guidelines by avoiding human/animal involvement and using public datasets with proper credit.
CRediT Author Statement: A standardized list describing each researcher’s role (e.g., writing, data collection). This paper’s authors used it to clarify contributions like “methodology” or “data analysis.”
Mendeley Data: A public repository where scientists share datasets. The radish leaf images are stored here, allowing other researchers to access and build on the work.
Open Access: Free availability of research online. This paper is open access, meaning anyone can read it without payment, promoting collaboration in agriculture and AI.
CC BY License: A Creative Commons license letting others share and adapt work, as long as they credit the original authors. This paper uses CC BY, encouraging widespread use of its dataset.
Sustainability in Agriculture: Farming practices that protect the environment and resources. By improving disease detection, this dataset helps reduce pesticide overuse and crop waste, supporting sustainability.
Food Security: Reliable access to affordable, nutritious food. Early disease detection in radishes (and other crops) ensures stable yields, which is critical for food security in regions like Bangladesh.
Smart Agriculture: Using technology (e.g., AI, sensors) to improve farming. This study contributes to smart agriculture by enabling automated disease diagnosis, helping farmers grow healthier crops.
Reference:
Barakala, M., Attada, V. R., Rajan, C., & Agnes, S. A. (2023). Rice plant leaf disease classification using deep residual learning. In 2023 International Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI) (pp. 1-8). IEEE. https://doi.org/10.1109/ICDSAAI59313.2023.10452139
