基于Mobilenetv3的果蔬分类识别研究

Research on Fruit and Vegetable Classification and Recognition based on Mobilenetv3

针对农贸市场、果蔬超市中结算流程不够智能化问题以及重型神经网络模型部署困难问题,基于轻量化神经网络模型Mobilenetv3,对果蔬分类识别进行了研究。首先针对果蔬超市、农贸市场环境复杂问题,提出了多样化数据采集方案,共采集果蔬170种,图片136000张。然后利用一系列增强方法,在训练时对数据进一步进行增强。最后使用训练数据集对Mobilenetv3进行了训练并使用测试数据集进行了测试,其top-1成功率达到了0.932,top-5成功率达到了0.991。研究结果表明:基于轻量化神经网络模型的果蔬分类可以用来辅助售货员进行果蔬分类。

Aiming at the problem of insufficient intelligence in the settlement process in the farmer’s market and fruit and vegetable supermarkets and the difficulty of deploying heavy-duty neural network models,based on the lightweight neural network model Mobilenetv3,this paper conducts a series of studies on the classification and recognition of fruits and vegetables.First,in response to the complex environment of fruit and vegetable supermarkets and farmer’s markets,a diversified data collection program was proposed,which collected 170 kinds of fruits and vegetables and 136,000 pictures.Then use a series of enhancement methods to further enhance the data during training.Finally,Mobilenetv3 was trained with the training data set and tested with the test data set.The top-1 success rate reached 0.932,and the top-5 success rate reached 0.991.Finally,the training data set was used to train Mobilenetv3 and the test data set was used to test.The top-1 success rate of the trained model reached 0.932,and the top-5 success rate reached 0.991.The research results show that the classification of fruits and vegetables based on the lightweight neural network model can be used to assist salespersons in the classification o f fruits and vegetables.