WalnutData: A UAV Remote Sensing Dataset of Green Walnuts and Model Evaluation

We conducted data collection on 8 walnut sample plots between July 18 and September 14, 2024. All these sample plots are located in Yangbi County, Dali Bai Autonomous Prefecture, Yunnan Province, China. In addition, in order to capture the changes in lighting conditions, we carried out the shooting between 9:00 and 19:00. The data collection equipment used uniformly was a DJI Matrice 300 RTK unmanned aerial vehicle (UAV) and a Zenmuse P1 (35mm F2.8) lens. The UAV took photos from a top-down angle (-90°) throughout the process according to the pre-planned flight route, and the flight path completely covered the scope of each sample plot. To reduce the impact of too high a flight altitude and too fast camera movement on the imaging quality, we set the flight speed between 1-3m/s and the flight altitude between 12-30m while ensuring flight safety.

Since the resolution of the UAV aerial images (8192×5460 pixels) is too large, which is not conducive to the training of the model, in this study, the screened original images were all cut with a step size of 512. The resolution of the cut images is 1024×1024 pixels. After the processing of the above steps, the dataset of this study was finally formed, which consists of a total of 30,240 images.