A Survey on the Pixel-Based Image Segmentation Techniques with the Intent of Optimization of System Throughput in the Field of Automation

Semantic segmentation, being one of the Computer Vision approaches, helps in portioning every object in an image, which makes it a good candidate that can be used in various applications like Medical Imaging and Autonomous vehicles. In the Medical field, clear analysis of images like MRI, CT scan, provides the accuracy in identifying the issue and to give proper medications. In the similar way, the pel-based segmentation (semantic segmentation) when used in identifying the objects on roads to ease the driving process requires computations in a precise way and often tend to use neural networks. These networks perform well when the size and amount of data is less but when it comes to training high resolution and larger amounts of data, it results in reduced flexibility of the desired implementation, thereby reducing its efficacy. In this work, the insights of methods and techniques used in the process of semantic segmentation in automation and the medical field is reviewed with the objective of providing a precise summary of techniques that can be used for selecting the best method to achieve best results. The findings of the comparison study shows that the COCO dataset has a large number of images and when used with efficient network architecture with criteria to segment, it can outperform. The contribution of this survey in the field of semantic segmentation outperforms by providing the precise description of deep learning techniques, listing out the complexities and associated techniques to reduce them, and summarizing associated datasets. comparison analysis of techniques to achieve accuracy with less computations.