State of the art object detection: A comparative study of YOLO and ViT

As trivial as it is for humans to identify, locate, and distinguish objects from their backgrounds and other objects, for computer systems, it is not so trivial. In Computer Science domain, the topic of Object detection addresses this issue. Object detection is one of the core tasks of computer vision, which lays the foundation for instance segmentation, semantic segmentation, panoptic segmentation etc. The last two decades have seen the tremendous evolution of object detection methods, from the Voila Jones detector to deep learning models to transformer-based models. The recent trends in the field are YOLO models and transformer-based models (such as DETR). We aim to explore object detection using YOLO and pure transformer ViT, do a comparative study, and furthermore, explore the theoretical approach of the hybrid model of YOLO and ViT.