Speed/accuracy trade-offs for modern convolutional object detectors

TL;DR

This paper systematically compares modern convolutional object detectors, analyzing how different architectures, feature extractors, and parameters affect the speed, accuracy, and memory usage to guide optimal selection for various applications.

Contribution

It provides a unified implementation of key detection architectures and maps out their speed-accuracy trade-offs, enabling better informed choices for different deployment scenarios.

Findings

Real-time detector suitable for mobile devices.

State-of-the-art accuracy achieved on COCO dataset.

Trade-off curves illustrating performance across architectures.

Abstract

The goal of this paper is to serve as a guide for selecting a detection architecture that achieves the right speed/memory/accuracy balance for a given application and platform. To this end, we investigate various ways to trade accuracy for speed and memory usage in modern convolutional object detection systems. A number of successful systems have been proposed in recent years, but apples-to-apples comparisons are difficult due to different base feature extractors (e.g., VGG, Residual Networks), different default image resolutions, as well as different hardware and software platforms. We present a unified implementation of the Faster R-CNN [Ren et al., 2015], R-FCN [Dai et al., 2016] and SSD [Liu et al., 2015] systems, which we view as "meta-architectures" and trace out the speed/accuracy trade-off curve created by using alternative feature extractors and varying other critical parameters such as image size within each of these meta-architectures. On one extreme end of this spectrum where speed and memory are critical, we present a detector that achieves real time speeds and can be deployed on a mobile device. On the opposite end in which accuracy is critical, we present a detector that achieves state-of-the-art performance measured on the COCO detection task.