UAV-WheatSeedling

Data Acquisition and Specifications​To facilitate this study, we constructed a custom UAV-based video dataset, designated as UAV-WheatSeedling, to capture winter wheat during the critical emergence stage. This dataset addresses the scarcity of high-quality, fine-grained UAV video data for winter wheat seedling phenotyping, a key bottleneck in precision agriculture. It supports the development and validation of computer vision models for seedling analysis under dynamic field conditions.​Data acquisition was conducted at the Anhui Academy of Agricultural Sciences, China (31.8877° N, 117.2460° E) from November to December 2024, covering the early emergence stage of winter wheat. Data collection included four scenarios: Sunny-Dry, Sunny-Wet, Overcast-Dry, and Overcast-Wet, to ensure model robustness against common environmental variations.​A DJI Air 2S drone was used to capture UHD video (3840×2160 pixels, 30 fps).​Data Content and File Specifications​The UAV-WheatSeedling dataset currently includes two files, with details as follows:​1. run.tar.gz: This file contains the model weights.​2. dataset.tar.gz: This file includes two parts:​history_frames: Historical frames prepared for temporal sequence analysis.​wheat_yolo_dataset: The dataset, which currently only includes the test set.