The ROC plot comparing the performance of trained CNN models under the Applied1 scenario in distinguishing between lensed and non-lensed samples. In this scenario, both the lensed and non-lensed populations have been altered compared to the Vanilla setting. The ensemble of EfficientNet-B3, EfficientNet-B4, and DenseNet-121 achieves a TP rate of 0.9112 with a FP rate of |$4.16 \times 10^{-4}$|⁠. The achieved TP rate surpasses that of both the Vanilla and Applied1 settings.