Traditional vine variety identification methods usually rely on the sampling of vine leaves followed by physical, physiological, biochemical and molecular measurement, which are destructive, time-consuming, labor-intensive and require experienced grape phenotype analysts. To mitigate these problems, this study aimed to develop an application (App) running on Android client to identify the wine grape automatically and in real-time, which can help the growers to quickly obtain the variety information. Experimental results showed that all Convolutional Neural Network (CNN) classification algorithms could achieve an accuracy of over 94% for twenty-one categories on validation data, which proves the feasibility of using transfer deep learning to identify grape species in field environments. In particular, the classification model with the highest average accuracy was GoogLeNet (99.91%) with a learning rate of 0.001, mini-batch size of 32 and maximum number of epochs in 80. Testing results of the App on Android devices also confirmed these results. Keywords: deep learning, mobile phone, grapevine cultivar, vine leaf image, identification, Vitis vinifera L. DOI: 10.25165/j.ijabe.20211405.6593 Citation: Liu Y X, Shen L, Su J Y, Lu N, Fang Y L, Liu F, et al. Development of a mobile application for identification of grapevine (Vitis vinifera L.) cultivars via deep learning. Int J Agric & Biol Eng, 2021; 14(5): 172–179.

Background Traditional vine variety identification methods usually rely on destructive sampling of vine leaves followed by physical, physiological, biochemical and molecular measurement, which is destructive, time-consuming, labor-intensive and requires experienced grape phenotype analysts. To mitigate these problems, this study aims to develop an application (App) running on Android client to identify wine grape automatically and in real time, which can help the growers to quickly obtain the variety information. Results Experimental results show that all Convolutional Neural Network (CNN) classification algorithms can achieve an accuracy of over 94% for 21 categories on validation data, which proves the feasibility of using transfer learning to identify grape species in field environments. In particular, the classification model with the highest average accuracy is Googlenet (97.4%) with learning rate (0.001), mini-batch size (32) and maximum number of epochs (80). Testing results of the App on Android device also confirms these results. Conclusions An App running on Android client is developed to identify the wine grape in real time and field condition, which can help the growers to quickly obtain the variety information of wine grape. A total of 4200 leaf images were first collected in the field environment, which contain 21 types of typical grapes. Then both image complement preprocessing and data augmentation are adopted to enhance image features and augment the training dataset. On this basis, a number of typical CNN models (including VGG-16, DenseNet, ResNet101, ResNet18, and GoogLeNet) are compared in transfer learning training to identify the suitable one while with model parameter tuning. It is shown that GoogLeNet model outperformed other models in terms of accuracy, model complexity, and robustness with a fine-tuned accuracy of 99.91%. The effect of image complement preprocessing is also assessed by using Grad-CAM algorithm. The developed App is also shown to be feasible for real-life applications with a processing time of less than 1 second.