Implementation and Comparative Analysis of Semantic Segmentation using Hybridized Deep Learning Models

Pixel-level object classification and delineation in computer vision is a crucial task that is addressed in this research, "Implementation and comparative analysis of semantic segmentation using hybridized deep learning models''. Convolutional neural nets (CNNs) have shown promise in semantic segmentation, but they still face difficulties when processing complicated visual situations. To take use of complementary features and improve segmentation accuracy, our novel method suggests fusing deep learning models by merging CNNs with Deep neural networks or attention mechanisms. The models such as mobilenetv2, vgg19, resnet152, inceptionnetv2, densenet169 are exploited for hybridization of the U-Net and the performance evaluation is done using aerial semantic segmentation drone dataset taken from Kaggle. Standard metrics such as pixel accuracy and mean intersection over union (mIoU) are used for evaluation.