Impact of Different Optimizers on InceptionResNet V2 for Amazon Fruit Classification

The Amazon rainforest is home to a diverse array of fruits, many of which have unique flavors and nutritional benefits. However, identifying and classifying these fruits can be challenging due to their morphological similarities and variations. Fruit categorization is a critical activity in a wide range of industrial applications. A fruit categorization system may assist a supermarket cashier in distinguishing between different fruit species and their costs. It may also be used to assist clients in determining if certain fruit species fit their nutritional requirements. This research paper provides a detailed classification of six different Amazon fruits, namely Acai, Cupuacu, Graviola, Guarana, Pupunha, and Tucuma, using a transfer learning model called InceptionResNet V2. The classification of these fruits can be challenging due to their morphological similarities and variations. To address this issue, various optimizers, including Adadelta, Adagrad, SGD, and RMSprop, were employed, and their performances were compared. Among these optimizers, the SGD optimizer yielded the best results, with an overall accuracy of 83%. The findings of this study will contribute to the existing literature on Amazon biodiversity and support efforts to promote sustainable production and conservation of these valuable resources.