2018年一种开放果园中芒果检测和计数方法的深度语义分割架构数据集

MangoNet数据集由定制的深度语义分割模型MangoNet收集，用于芒果果实检测。该数据集由49张jpg格式的高分辨率4000×3000彩色图像组成。上述图像在自然光照条件下的芒果园中采集，其中45幅图像用于训练，4幅图像用于测试。研究者为每个图像进行了水果和非水果类像素级注释。如需使用数据集训练深度学习模型，用户需要裁剪原始的大数据图像，分割成小块，例如MangoNet的200×200像素大小，以规避计算内存限制问题。用户可通过使用与所需模型输入图像大小相同的滑动窗口对整个图像进行采样完成图像裁剪。https://github.com/avadesh02/MangoNet-Semantic-Dataset

The MangoNet dataset was collected by the custom deep semantic segmentation model MangoNet for mango fruit detection. It consists of 49 high-resolution 4000×3000 color images in JPEG format. All images were captured in mango orchards under natural lighting conditions, with 45 images designated for training and 4 for testing. Researchers performed pixel-level annotations for every image, categorizing pixels into fruit and non-fruit classes. If users wish to train deep learning models with this dataset, they must crop the original high-resolution images into smaller patches, such as the 200×200 pixel size used in MangoNet, to avoid computational memory limitations. Users can complete the cropping by sampling the entire image via a sliding window that matches the input image size required by the target model. https://github.com/avadesh02/MangoNet-Semantic-Dataset