<![CDATA[Molly fish (Poecilia sphenops) is one of the ornamental fish species that is widely cultured. This study aims to develop a classification system for ornamental molly fish using the VGG16 model, trained with on-the-fly data augmentation techniques (flip, zoom, rotation, and translation). The dataset used consists of 1,750 images of molly fish, divided into seven different species: Black, Blue Electric, Calico, Dalmatian, Golden Black, Platinum, and Sunkist. Data augmentation is performed dynamically during the training process without saving the transformation results, aiming to increase data diversity and help the model recognize patterns more accurately. The experimental results show that the optimal combination of parameters, namely a learning rate of 1e-5, a batch size of 32, and 50 epochs, achieved a training accuracy of 97.80%, validation accuracy of 99.61%, and test accuracy of 99.62%. Additionally, very high precision (99.63%), recall (99.62%), and F1-Score (99.62%) values were achieved. Although there were minor classification errors in the "Black" class predicted as "Sunkist," these errors were minimal and did not affect the overall results. This study shows that with the right parameter settings and the use of augmentation techniques, the VGG16 model can provide classification results with fairly high accuracy for molly ornamental fish. This model also has the potential to be applied in the ornamental fish aquaculture industry, particularly in image-based automatic detection systems.]]>
Copyrights © 2025
