VisionFusion (Vis-Fusion): Integrating CNNs and Vision Transformers for Superior Object Detection

Object detection has long been a critical and challenging problem in the field of computer vision. Recently, Vision Transformers (ViT) have demonstrated exceptional performance in image classification tasks, highlighting their potential for other vision applications. In this paper, we introduce Vis-Fusion, an innovative object detection model based on Vision Transformers. By leveraging the strengths of the self-attention mechanism and convolutional neural networks, along with the integration of the GCCA (Global Context Block and Coordinate Attention) module, Vis-Fusion significantly enhances both the accuracy and robustness of object detection. We evaluate the performance of Vis-Fusion on the Mini COCO benchmark dataset, and our experimental results show that Vis-Fusion outperforms existing state-of-the-art object detection methods in terms of both detection accuracy and processing speed. This study highlights the promising potential of Transformer-based architectures for complex visual tasks and provides valuable insights for future research in this area.