Simple Training Strategies and Model Scaling for Object Detection

TL;DR

This paper evaluates various training and scaling strategies for object detection, demonstrating significant accuracy improvements and proposing simple model scaling methods to optimize the speed-accuracy trade-off.

Contribution

It systematically benchmarks training techniques, introduces simple scaling strategies for detection models, and compares backbone architectures, highlighting effective methods for improving object detection performance.

Findings

Vanilla detectors improved by 7.7% in accuracy and 30% faster.

Introduced RetinaNet-RS and Cascade RCNN-RS scaling strategies.

ResNet with minor modifications outperforms EfficientNet as backbone.

Abstract

The speed-accuracy Pareto curve of object detection systems have advanced through a combination of better model architectures, training and inference methods. In this paper, we methodically evaluate a variety of these techniques to understand where most of the improvements in modern detection systems come from. We benchmark these improvements on the vanilla ResNet-FPN backbone with RetinaNet and RCNN detectors. The vanilla detectors are improved by 7.7% in accuracy while being 30% faster in speed. We further provide simple scaling strategies to generate family of models that form two Pareto curves, named RetinaNet-RS and Cascade RCNN-RS. These simple rescaled detectors explore the speed-accuracy trade-off between the one-stage RetinaNet detectors and two-stage RCNN detectors. Our largest Cascade RCNN-RS models achieve 52.9% AP with a ResNet152-FPN backbone and 53.6% with a SpineNet143L backbone. Finally, we show the ResNet architecture, with three minor architectural changes, outperforms EfficientNet as the backbone for object detection and instance segmentation systems.