Aquafarming faces unique challenges that certainly affect effective management of aquafarming, which includes small species like prawn, shrimp, larvae of these species. Depending on the quality of water including low-light conditions, occlusions, variations in image quality, predation, feeds, breeding condition, environmental pollution and climatic change, the productivity, sustainability and profitability in culturing small species is highly affected. Henceforth, the accurate identification of small species in aquafarming is essential. This research work explores a cutting-edge approach that leverages State-of-the-Art (SOTA) deep learning models which improves the accuracy of species identification in aquafarming. Leveraging the benefits of a custom dataset, that encompasses size variation, morphology including different environment conditions, the study provides a comparison of various methodology that integrates advanced Convolutional Neural Network (CNN) models like ResNet50, DenseNet201, EfficientNetB2, MobileNet_V2, and InceptionResNetV2 as backbone which extract the feature and classify, analyze and detect small aquafarming species. The work evaluates the performance of these models by focusing on performance metrics like accuracy, precision, recall, F1 score, and ROC Curve. On training these models for 10 Epochs, Resnet50 and DenseNet201 give 98.55% of validation accuracy in 23.27 min and 10.73 min respectively. This work contributes to the growing field of aquafarming by offering a scalable and efficient solution for selecting the appropriate model which provides effective small aquafarming species identification and classification, reducing labor costs, and enabling real-time monitoring.