Recognition of grape leaf diseases using MobileNetV3 and deep transfer learning

Timely diagnosis and accurate identification of grape leaf diseases are decisive for controlling the spread disease ensuring healthy development industry. The objective this research was to propose a simple efficient approach improve accuracy with limited computing resources scale training image dataset based on deep transfer learning an improved MobileNetV3 model (GLD-DTL). A pre-training obtained by using ImageNet extract common features leaves. And last convolution layer modified adding batch normalization function. dropout followed fully connected used generalization ability realize weight matrix quantify scores six diseases, according which Softmax method added as top networks give probability distribution diseases. Finally, dataset, constructed processing data augmentation annotation technologies, input into retrain obtain recognition (GLDR) model. Results showed that proposed GLD-DTL had better performance than some recent approaches. high 99.84% while size small 30 MB. Keywords: real-time recognition, learning, DOI: 10.25165/j.ijabe.20221503.7062 Citation: Yin X, Li W H, Z, Yi L L. Recognition learning. Int J Agric & Biol Eng, 2022; 15(3): 184–194.