Simplified-Pix2Seq: A Simple Language Modeling Framework for Vehicle detection

With the development of intelligent transportation systems and the rise of autonomous driving, vehicle detection and classification have become increasingly important. Although Transformer-based object detection models[1] have achieved high accuracy, existing methods are often complex and rely on large-scale datasets, making them unsuitable for fast and efficient traffic scene detection. In this paper, we propose an improved Simplified-Pix2Seq model based on Google’s Pix2Seq [2], a simple, effective end-to-end detection framework—applied to vehicle detection and type recognition in traffic scenes. Experimental results show that the model maintains a balance between detection accuracy and speed while keeping low complexity, effectively addressing detection challenges in traffic scenarios.