Improved segmentation model for OC detection using deep ensemble architecture

This study introduces an innovative deep ensemble architecture, ensemble of DBN, SqueezeNet, and IDCNN (en-IDCNN) models for OC detection with improved DeepJoint segmentation model via dual input. This model adheres to a structured procedure encompassing pre-processing, segmentation, feature extraction, feature level fusion, and detection. Input data and image undergoes individual pre-processing. Subsequently, improved DeepJoint segmentation is utilised to the pre-processed image. Both pre-processed data and segmented images retrieving the feature independently through feature extraction. Afterward, the retrieved features are fused through feature level fusion. The fused features are then processed independently using deep belief network (DBN), SqueezeNet, and improved deep convolutional neural network (IDCNN) for detecting the OC. Finally, the scores attained from these three models are combined using score level fusion. To confirm the proposed en-IDCNN model, thorough examines encompassing simulation and experimental calculations are carried out.