Xception Net & Vision Transformer: A comparative study for Deepfake Detection

Enhancing Retinal Scan Classification: A Comparative Study of Transfer Learning and Ensemble Techniques

“…Similar to risk classification for multidisease classification the Ensemble 2 performs best with metric scores of (accuracy = 76.92%, f1-score = 0.7128, precision = 0.5579, recall = 0.7063). The models taken for ensemble systems are: XceptionNet [26], Resnet152V2 [25] and EfficientNetV2M [27] as they are the top performers.…”

“…This technique has been used for both risk classification and multi disease classification. The models used during training were DenseNet 201 [24], ResNet152V2 [25], XceptionNet [26], EfficientNetV2 (S, M, L) [27], EfficientNetB7 [28], MobileNetV2 [29]. Each of these models were trained for 30 epochs.…”

“…The scan is then passed to the selected models. The models used in the final ensemble system are XceptionNet [26], ResNet152V2 [25] and EfficientNetV2M [27]. These models are selected based on their metric scores (accuracy, f1-score, precision, recall).…”

“…April 2023 Revised:18 May 2023 Accepted: 08 June 2023 ___________________________________________________________________________________________________________________ 96.98%, f1-score = 0.9207, precision = 0.8757, recall = 0.9135). The models taken for ensemble systems are: XceptionNet[26], Resnet152V2[25] and EfficientNetV2M[27], as they are the top performers. Both the ensemble methods show an improvement over the performance of individual models.…”