Comparative Study of TRANS - GAN Architecture for Bio-Medical Image Semantic Segmentation
Malvika Ashok, Abhishek Gupta · 2022 7th International Conference on Communication and Electronics Systems (ICCES) · 2022
In the area of biomedical image processing, medical image segmentation plays a crucial role. Today due to the deep sculptures of deep neural networks and innovative by-passes like the Transformers this field has rejuvenated. This paper analyzes various algorithmic advancements regarding Transformers and Generative Adversarial Networks (GAN) which have paved the anatomy of image segmentation. The GAN network is not only able to regenerate feature spaces but also can transpose the inputs from the transformer. These algorithms are not only at par with the bifurcation of networks like U-net and ResNet but showcase a strong and more intuitive idea of human neural intelligence. Finally, all methods are compared based on accuracy, dice score, architectural evaluation, etc. The ability of these networks are deduced to regenerate and recreate the feature spaces for the given input image and thus can segment high features for prediction.