Shallow Networks for High-Accuracy Road Object-Detection

TL;DR

This paper demonstrates that simple modifications like increasing input resolution and using shallow CNN layers can significantly improve road object detection accuracy, challenging the trend of using deeper models.

Contribution

It reveals that shallow networks combined with higher resolution inputs can achieve high accuracy, offering a simpler alternative to complex deep models.

Findings

Upsampling improves accuracy by up to 12 percentage points.

Shallow CNN layers can outperform deeper layers in certain scenarios.

Shallow models with high-resolution images are competitive with deep models.

Abstract

The ability to automatically detect other vehicles on the road is vital to the safety of partially-autonomous and fully-autonomous vehicles. Most of the high-accuracy techniques for this task are based on R-CNN or one of its faster variants. In the research community, much emphasis has been applied to using 3D vision or complex R-CNN variants to achieve higher accuracy. However, are there more straightforward modifications that could deliver higher accuracy? Yes. We show that increasing input image resolution (i.e. upsampling) offers up to 12 percentage-points higher accuracy compared to an off-the-shelf baseline. We also find situations where earlier/shallower layers of CNN provide higher accuracy than later/deeper layers. We further show that shallow models and upsampled images yield competitive accuracy. Our findings contrast with the current trend towards deeper and larger models to achieve high accuracy in domain specific detection tasks.