Object Detection in 20 Years: A Survey

This article provides an extensive survey of object detection, one of the most fundamental and challenging problems in computer vision. It reviews the field in light of its technical evolution over more than a quarter-century, from the 1990s to 2022, contrasting today's deep-learning-driven detection techniques with the ingenious design thinking of early computer vision. The survey systematically organizes the field by covering milestone detectors in history, detection datasets, evaluation metrics, and the fundamental building blocks of detection systems.

Beyond cataloguing historical methods, the article also addresses speedup techniques and recent state-of-the-art detection methods, giving readers a coherent picture of how the field has progressed. By spanning such a long time horizon and connecting early perspective design to the deep learning revolution, it serves as a reference point for understanding both the trajectory and the profound impact object detection has had on the entire computer vision field.