基于卷积神经网络的食物识别及实现

Convolutional neural network based food recognition and implementation

在超市无人结算服务中,使用电子标签对部分货物如水果、蔬菜等进行结算的成本过高、便捷性不高,至今依然采用人工结算的方式。针对这一问题,本文提出了基于卷积神经网络的食物识别方法。通过自建水果数据集来训练卷积神经网络分类模型;基于训练后的模型构建可视化平台进行食物识别。实验结果表明,利用卷积神经网络的食物识别的预测准确率为96.34%。

In the unmanned settlement service of supermarkets,the cost of using electronic tags to settle some goods such as fruits and vegetables is too high and the convenience is not high,and manual settlement is still used so far.To address this problem,this paper proposes a food recognition method based on convolutional neural network.A self-built fruit dataset is used to train a convolutional neural network classification model;a visualization platform is constructed based on the trained model for food recognition.The experimental results show that the prediction accuracy of food recognition using convolutional neural network is 96.34%.