Resultant force on grains of a real sand dune

This repository contains the dataaset used to train a convolutional neural network (CNN) designed to measure the resultant forces acting on a barchan dune. The model has been trained using numerical data to predict force distributions based on the morphological features of dunes. In our research, we developed a groundbreaking method that leverages deep learning to estimate the forces acting on real barchan dunes at grain scale. This method combines image data from subaqueous experiments with numerical simulations, offering a novel approach to granular mechanics measurement. The trained CNN model is capable of predicting force distributions on dunes from their morphological features, even when applied to dune configurations not seen during training.本仓库包含用于训练卷积神经网络（CNN）的数据集，该网络旨在测量作用于新月形沙丘（barchan dune）的合力。本模型通过数值数据训练完成，可基于沙丘的形态学特征预测受力分布。 在本研究中，我们提出了一种突破性方法，借助深度学习以颗粒尺度估算真实新月形沙丘所受的作用力。该方法将水下实验获取的图像数据与数值模拟相结合，为颗粒力学测量提供了全新的研究路径。经训练完成的卷积神经网络模型，可仅通过沙丘的形态学特征预测其受力分布，即便将其应用于训练阶段未见过的沙丘构型时依然有效。