DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation

TL;DR

DFANet is a highly efficient CNN architecture designed for real-time semantic segmentation, balancing speed and accuracy by reducing parameters and FLOPs while maintaining strong performance on benchmark datasets.

Contribution

The paper introduces DFANet, a lightweight CNN with multi-scale feature aggregation that achieves state-of-the-art real-time segmentation performance with significantly fewer computational resources.

Findings

Achieves 70.3% Mean IOU on Cityscapes with 1.7 GFLOPs

Runs at 160 FPS on a Titan X GPU

Uses 8× fewer FLOPs than previous methods

Abstract

This paper introduces an extremely efficient CNN architecture named DFANet for semantic segmentation under resource constraints. Our proposed network starts from a single lightweight backbone and aggregates discriminative features through sub-network and sub-stage cascade respectively. Based on the multi-scale feature propagation, DFANet substantially reduces the number of parameters, but still obtains sufficient receptive field and enhances the model learning ability, which strikes a balance between the speed and segmentation performance. Experiments on Cityscapes and CamVid datasets demonstrate the superior performance of DFANet with 8$\times$ less FLOPs and 2$\times$ faster than the existing state-of-the-art real-time semantic segmentation methods while providing comparable accuracy. Specifically, it achieves 70.3\% Mean IOU on the Cityscapes test dataset with only 1.7 GFLOPs and a speed of 160 FPS on one NVIDIA Titan X card, and 71.3\% Mean IOU with 3.4 GFLOPs while inferring on a higher resolution image.