A collection of approximately 500 annotated images of cocoa trees, acquired in Côte d'Ivoire in 2019. It was created to evaluate neural network performance for detecting cocoa pods as part of the European Cocoa4Future project.
Use Cases
- Train and evaluate object detection neural networks on annotated cocoa pod images.
- Benchmark model performance for agricultural yield estimation using the provided image annotations.
- Analyze visual features of cocoa pods in field conditions from the 2019 Côte d'Ivoire imagery.
Strengths
- Approximately 500 annotated images provide a foundation for model training.
- Images have specific geographic (Côte d'Ivoire) and temporal (2019) context.
- Created for a defined research purpose within the European Cocoa4Future project.
Limitations
- The dataset size of ~500 images is relatively small for training deep neural networks without augmentation.
- Annotations and image quality are unknown and cannot be verified from the description.
- Data is from a single year and country, limiting generalizability to other regions or seasons.
Provenance
- Source
- CIRAD Harvested Collection, author Borianne Philippe.
- Collection Method
- Images were acquired in Côte d'Ivoire as part of the European Cocoa4Future project.
- Time Range
- 2019
- Freshness
- null
- Geography
- Côte d'Ivoire