PlantVillage
PlantVillage is a reference dataset in digital agriculture, specialized in the detection of plant diseases. It brings together thousands of images of crop leaves annotated according to their health status, making it a very useful tool for AI-based agricultural diagnostic support systems.
Approximately 54,000 images in JPEG format
Free for academic research. For commercial use, consult the specific conditions of the license
Description
The PlantVillage dataset includes:
- 54,306 JPEG images of plant leaves
- Annotations such as classification by culture and pathology
- 38 classes covering crop-disease combinations (e.g. tomato — mildew, potato — brown spot, etc.)
The images are taken on a neutral background and under controlled conditions, which allows standardization useful for the initial training of models, with the possibility of fine-tuning on data in real conditions.
What is this dataset for?
PlantVillage is used for:
- Training models for the recognition of plant diseases using images
- The creation of mobile agricultural diagnostic applications
- The development of precision agriculture and crop monitoring tools
- The improvement of alert and phytosanitary prevention systems
Can it be enriched or improved?
Yes, PlantVillage can be optimized by:
- The addition of images in natural conditions (terrain, variable lighting, non-neutral background)
- The integration of contextual data such as location, climate or season
- The addition of videos or time series to detect disease progression
- Combining with other open source sources like PlantDoc or AgriNet for a more robust model
🔗 Source: PlantVillage Dataset
Frequently Asked Questions
Can PlantVillage be used in production in agricultural applications?
Yes, after fine-tuning on real data. The dataset is very useful in the prototyping phase, but adapting to local conditions is essential for optimal performance in production.
Does PlantVillage cover rare or emerging diseases?
Not directly. The dataset focuses on the most common pathologies. For emerging or regional diseases, it may be necessary to collect new data or to use specialized extensions.
Can PlantVillage be used with embedded sensors or drones?
Yes, but a rehabilitation step is required: the PlantVillage images being in a neutral background, it is important to reconstitute a mixed dataset with images taken in real context to generalize the models to aerial or outdoor views.
