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

TLDR

Real‑time pixel‑wise semantic segmentation is crucial for mobile applications, yet existing deep neural networks suffer from high computational cost and long run‑times. The paper proposes ENet, a novel deep neural network architecture designed for low‑latency semantic segmentation. ENet was evaluated on CamVid, Cityscapes, and SUN datasets, with performance measured on embedded systems and comparisons to state‑of‑the‑art methods, highlighting accuracy‑time trade‑offs and potential software optimizations. ENet achieves up to 18× speed‑up, 75× fewer FLOPs, 79× fewer parameters, while matching or surpassing the accuracy of existing models.

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.