Search

In an agricultural field, plant phenotyping using object detection models is gaining attention. However, collecting the training data necessary to create generic and high-precision models is extremely challenging due to the difficulty of annotation and the diversity of domains. Furthermore, it is difficult to transfer training data across different crops, and although machine learning models effective for specific environments, conditions, or crops have been developed, they cannot be widely applied in actual fields. In this study, we propose a generative data augmentation method (D4) for vineyard shoot detection. D4 uses a pre-trained text-guided diffusion model based on a large number of original images culled from video data collected by unmanned ground vehicles or other means, and a small number of annotated datasets. The proposed method generates new annotated images with background information adapted to the target domain while retaining annotation information necessary for object detection. In addition, D4 overcomes the lack of training data in agriculture, including the difficulty of annotation and diversity of domains. We confirmed that this generative data augmentation method improved the mean average precision by up to 28.65% for the BBox detection task and the average precision by up to 13.73% for the keypoint detection task for vineyard shoot detection. Our generative data augmentation method D4 is expected to simultaneously solve the cost and domain diversity issues of training data generation in agriculture and improve the generalization performance of detection models.

We present an analysis of the impact of heterogeneous chemistry on the partitioning of nitrogen species measured by the Upper Atmosphere Research Satellite (UARS) instruments. The UARS measurements utilized include N2O, HNO3, and ClONO2 from the cryogenic limb array etalon spectrometer (CLAES), version 7 (v.7), and temperature, methane, ozone, H2O, HCl, NO and NO2 from the halogen occultation experiment (HALOE), version 18. The analysis is carried out for the UARS data obtained between January 1992 and September 1994 in the 100-to 1-mbar (approx. 17-47 km) altitude range and over 10 degrees latitude bins from 70 S to 70 N. The spatiotemporal evolution of aerosol surface area density (SAD) is adopted from analysis of the Stratospheric Aerosol and Gas Experiment (SAGE) II data. A diurnal steady state photochemical box model, constrained by the temperature, ozone, H2O, CH4, aerosol SAD, and columns of O2 and O3 above the point of interest, has been used as the main tool to analyze these data. Total inorganic nitrogen (NOy) is obtained by three different methods: (1) as a sum of the UARS-measured NO, NO2, HNO3, and ClONO2; (2) from the N2O-NOy correlation, and (3) from the CH4-NOy correlation. To validate our current understanding of stratospheric heterogeneous chemistry for post-Pinatubo conditions, the model-calculated monthly averaged NOx/NOy ratios and the NO, NO2, and HNO3 profiles are compared with the UARS-derived data. In general, the UARS-constrained box model captures the main features of nitrogen species partitioning in the post-Pinatubo years, such as recovery of NOx after the eruption, their seasonal variability and vertical profiles. However, the model underestimates the NO2 content, particularly in the 30- to 7-mbar (approx.23-32 km) range. Comparisons of the calculated temporal behavior of the partial columns of NO2 and HNO3 and ground-based measurements at 45 S and 45 N are also presented. Our analysis indicates that ground-based and HALOE v.18 measurements of the NO2 vertical columns are consistent within the range of their uncertainties and are systematically higher (up to 50%) than the model results at midlatitudes in both hemispheres. Reasonable agreement is obtained for HNO3 columns at 45 S, suggesting some problems with nitrogen species partitioning in the model. Outstanding uncertainties are discussed.