AgriVision-Peach: A Computer Vision Dataset for Peach Fruits in Intelligent Thinning

AgriVision-Peach is a real peach orchard scene image dataset designed specifically for automated fruit thinning operations. Combining computer vision and artificial intelligence technologies, it provides high-quality training data for peach fruit recognition, localization, and counting in precision agriculture, supporting the development of intelligent fruit thinning systems and the application of agricultural robots.

Peach fruit thinning is a key agronomic operation. Manual thinning is labor-intensive and inefficient; accurate counting and evaluation are challenging when fruits are dense. Automated thinning has emerged as a research focus, but agricultural scenarios present visual recognition challenges like significant lighting variations, fruit occlusion, and similar color backgrounds, making general models difficult to apply directly.

This dataset is built by the MachineLearningVisionRG team, aiming to provide training data for automated fruit thinning systems. All images are from real peach orchards, covering different growth stages, lighting conditions, and leaf/fruit occlusion situations. The diverse scenarios enhance the generalization ability of models.

The dataset can support tasks such as object detection (identifying fruit positions), instance segmentation (outlining fruit contours), and fruit counting (quantifying fruit-bearing status). It has direct value for intelligent fruit thinning robots: the robot cruises autonomously, analyzes fruit-bearing status in real time, guides the robotic arm for precise operations, improves efficiency, and optimizes decision-making.

AgriVision-Peach fills the gap in the subfield of agricultural computer vision. Compared to general datasets, it takes into account practical factors such as maturity, pests and diseases, and variety differences, which are closer to application needs. High-quality datasets are the foundation of data-driven decision-making; open sharing promotes academic exchanges and the iteration of agricultural AI.

In the future, we can explore multi-modal data fusion (visible light + depth + spectrum), time-series data analysis (growth model to predict thinning timing), and cross-variety transfer learning. With the decline in edge computing and robot costs, intelligent fruit thinning systems are expected to be applied on a large scale, and the dataset provides important data infrastructure.