YOLO -- You only look 10647 times

TL;DR

This paper explains the YOLO object detection method as a parallel classification of over ten thousand fixed region proposals, linking it conceptually to other detection and classification models, and provides visualization tools for better understanding.

Contribution

It offers a new perspective on YOLO as a parallel classification process and introduces visualization tools to explore YOLO's information processing streams.

Findings

YOLO's output pixels attend to specific sub-regions, acting like region proposals.

This view unifies YOLO with RCNN and ResNet models conceptually.

Visualization tools enhance understanding of YOLO's internal processing.

Abstract

With this work we are explaining the "You Only Look Once" (YOLO) single-stage object detection approach as a parallel classification of 10647 fixed region proposals. We support this view by showing that each of YOLOs output pixel is attentive to a specific sub-region of previous layers, comparable to a local region proposal. This understanding reduces the conceptual gap between YOLO-like single-stage object detection models, RCNN-like two-stage region proposal based models, and ResNet-like image classification models. In addition, we created interactive exploration tools for a better visual understanding of the YOLO information processing streams: https://limchr.github.io/yolo_visualization