| dc.description.abstract | Orange (Citrus spp.) is one of the cultivated fruits in Indonesia, with a production reaching 2,551,999.00 tons in 2022 according to the Central Statistics Agency (BPS). Oranges are highly popular for consumption due to their rich nutritional content. Besides providing essential nutrients and energy, oranges are considered valuable sources of nutrition and health supplements. The quality taste of oranges plays a crucial role in the sustainability of the orange cultivation industry, influenced by factors such as orange varieties, harvesting seasons, cultivation methods, and environmental factors. The taste of an orange is hard to determine without damaging the fruit. This can lead to waste when discarding oranges that don't taste good after being sampled. EfficientNet architecture is a type of Convolutional Neural Network (CNN) architecture that intelligently combines scaling techniques such as width, depth, and image resolution. In this study, a dataset consisting of 608 data points was utilized, divided into 424 data points for training, 120 for validation, and 64 for testing. The data underwent pre-processing stages, including resizing, cropping, flipping, as well as rotating images by 45 and 90 degrees. After pre-processing, the data was fed into the Convolutional Neural Network (CNN) algorithm with EfficientNet-B4 architecture. The training process of the model involved 40 epochs with a batch size of 30. The results of using the CNN algorithm with EfficientNet-B4 architecture showed an accuracy rate of 96.88%. | en_US |
