Prior aquatic animal image classification research focused on distinguishing external features in controlled settings, utilizing either digital cameras or webcams. Identifying visually similar species, like Short mackerel (Rastrelliger brachysoma) and Indian mackerel (Rastrelliger kanagurta), is challenging without specialized knowledge. However, advancements in computer technology have paved the way for leveraging machine learning and deep learning systems to address such challenges. In this study, transfer learning techniques were employed, utilizing established pre-trained models such as ResNet50, Xception, InceptionV3, VGG19, VGG16, and MobileNetV3Small. These models were applied to differentiate between the two species using raw images captured by a smartphone under uncontrolled conditions. The core architecture of the pre-trained models remained unchanged, except for the removal of the final fully connected layer. Instead, a global average pooling layer and two dense layers were appended at the end, comprising 1024 units and by a single unit, respectively. To mitigate overfitting concerns, early stopping was implemented. The results revealed that, among the models assessed, the Xception model exhibited the most promising predictive performance. It achieved the highest average accuracy levels of 0.849 and 0.754 during training and validation, surpassing the other models. Furthermore, fine-tuning the Xception model by extending the number of epochs yielded more impressive outcomes. After 30 epochs of fine-tuning, the Xception model demonstrated optimal performance, reaching an accuracy of 0.843 and displaying a 11.508% improvement in predictions compared to the model without fine-tuning. These findings highlight the efficacy of transfer learning, particularly with the Xception model, in accurately distinguishing visually similar aquatic species using smartphone-captured images, even in uncontrolled conditions.

This study proposed an artificial intelligence framework for red tilapia weight-range classification from UAV-based aerial imagery using two hybrid deep learning models, namely Hybrid CNN-XGBoost and Hybrid EfficientNet-B0-XGBoost. Two image sizes were evaluated: 5 × 5 m images, which preserved the spatial context of the cage culture system, and 2 × 2 m images, which focused on areas with high fish aggregation density. For the 5 × 5 m images, Hybrid CNN-XGBoost achieved the highest performance, with a mean accuracy of 0.988 ± 0.008 (98.8%) at 20 tuning units, whereas Hybrid EfficientNet-B0-XGBoost achieved 0.977 ± 0.021 (97.7%) at 40 tuning units. In addition, Hybrid CNN-XGBoost exhibited a substantially shorter average computational workflow time per image (0.038 ± 0.001 s) than Hybrid EfficientNet-B0-XGBoost (65.007 ± 6.141 s). In contrast, under the 2 × 2 m image condition with limited spatial context, Hybrid EfficientNet-B0-XGBoost outperformed Hybrid CNN-XGBoost, achieving the highest mean accuracy of 0.900 ± 0.010 (90.0%) at 30 tuning units, compared with 0.850 ± 0.022 (85.0%) at 50 tuning units. These findings indicate that larger images improve classification accuracy by preserving spatial context, whereas EfficientNet-B0 enhances deep feature extraction capability and improves classification accuracy when spatial context is limited.