ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation

TL;DR

ENet is a novel deep neural network architecture optimized for real-time semantic segmentation, achieving significant speed and efficiency improvements while maintaining comparable accuracy to existing models.

Contribution

The paper introduces ENet, a lightweight neural network architecture designed specifically for low-latency semantic segmentation tasks, with substantial reductions in FLOPs, parameters, and processing time.

Findings

ENet is up to 18× faster than existing models.

Requires 75× fewer FLOPs and 79× fewer parameters.

Achieves similar or better accuracy on benchmark datasets.

Abstract

The ability to perform pixel-wise semantic segmentation in real-time is of paramount importance in mobile applications. Recent deep neural networks aimed at this task have the disadvantage of requiring a large number of floating point operations and have long run-times that hinder their usability. In this paper, we propose a novel deep neural network architecture named ENet (efficient neural network), created specifically for tasks requiring low latency operation. ENet is up to 18$\times$ faster, requires 75$\times$ less FLOPs, has 79$\times$ less parameters, and provides similar or better accuracy to existing models. We have tested it on CamVid, Cityscapes and SUN datasets and report on comparisons with existing state-of-the-art methods, and the trade-offs between accuracy and processing time of a network. We present performance measurements of the proposed architecture on embedded systems and suggest possible software improvements that could make ENet even faster.