Deep learning optimizer performance analysis for pomegranate fruit quality gradation

Quality and safety are important factors in the food industry. In recent years automatic visual inspection technology has become more potential and important for fruit grading applications. This is because quality is an important factor for consumers and so essential for the market. This paper focuses on a comparative study of deep learning optimizers for pomegranate fruit quality grading. It plays an important role in maximizing the efficiency of the neural network model. Optimizers are mathematical functions or algorithms which are dependent on various parameters of the model i.e., weights and biases. This paper presents the performances of the various deep learning optimizers for pomegranate fruit quality grading. The dataset used for this study is named as Pomegranate Fruit dataset from the Kaggle dataset. Dataset has three grades G1, G2, and G3. Each grade is having four internal quality labels and has 90 images in it. Training is done using SGD, Adadelta, Adagrad, RMSprop, and Adam optimizers. This study helped in analyzing better optimizer and identifying the need for overall improvement in performance of the optimization.