DRAGONS

Grapes Ready for Picking: Investigating the Readiness of Precision Viticulture Datasets for Scene Reconstruction

Michał Jan Włodarczyk, Katarzyna Wasielewska-Michniewska, Przemysław Hołda, Maria Ganzha, Marcin Paprzycki, Nico Heider, Martin Schieck, Bogdan Franczyk, Tasuku Makabe, Iori Yanokura, Kei Okada

Sensors, 26(21), 6708, 2026

Public datasets of vineyard imagery were collected for detection, segmentation and disease mapping and are now proposed as a source for 3D scene reconstruction. Their documentation rarely states the properties that reconstruction depends on, such as view overlap, parallax and frame spacing, so it is unclear which can be reconstructed. This study measures that for the publicly available vineyard datasets. A structured search identifies eighteen datasets. Ten are characterized under a schema of thirty-six documented and measured properties, and only six contain overlapping views. Every sequence that registers is reconstructed with one fixed pipeline of structure from motion and 3D Gaussian splatting, giving twenty-seven reconstructions on two held-out splits. Registration is decided by the capture rather than by the dataset; novel-view quality varies as much within a dataset as between datasets; and frame spacing is the one capture property that limits reconstruction. Three preprocessing steps are tested as ablations on the failure cases. Band alignment restores registration, whereas dehazing a backlit pass and masking the chassis of a ground robot give no improvement that this study can resolve. The schema and pipeline apply to any vineyard dataset, and the results are published as an open online catalog.

Practical Considerations and Challenges for Adaptation and Transferability of UAV and AI-based Precision Viticulture

Katarzyna Wasielewska-Michniewska, Przemysław Hołda, Michał Jan Włodarczyk, Maria Ganzha, Marcin Paprzycki, Nico Heider, Martin Schieck, Bogdan Franczyk, Tasuku Makabe, Iori Yanokura, Kei Okada, Barbara Wiza

In recent years, automation and optimization in the domain of Precision Viticulture (PV) have motivated extensive research on AI-based UAV-supported methods for tasks such as yield estimation, disease detection, field monitoring, and spraying. The literature includes evidence from simulation and real-world deployments that demonstrate the benefits of PV solutions. However, the transferability across sites and generalization of the proposed solutions still remain open challenges. Among others, even though numerous public datasets are available and methods have been identified to address key problems, PV systems are usually targeted at a single deployment. While existing review papers primarily focus on technological developments, application scenarios, and methodological advances in PV, UAV systems, and AI, comparatively less attention has been devoted to the practical factors affecting the transferability of these solutions across different vineyard environments. To address this gap, this paper adopts a deployment-oriented perspective, examining the key factors and considerations that influence the transferability and generalization of AI-based UAV solutions in PV. Rather than reviewing individual methods solely in terms of their performance, the contribution focuses on identifying practical challenges and considerations relevant to assessing and improving their applicability across different deployment scenarios.

Literature library