EKENet: Efficient knowledge enhanced network for real-time scene parsing

作者：

Highlights：

• We propose an Efficient Dual Abstraction block, which is able to extract rich features with low computation complexity.

• We introduce a novel light-weight Encoding-Enhancing module to enhance the representation of high-resolution feature map.

• Our fully-equipped model (EKENet), achieves the new state-of-the-art performance in terms of speed and accuracy tradeoff.

摘要

•We propose an Efficient Dual Abstraction block, which is able to extract rich features with low computation complexity.•We introduce a novel light-weight Encoding-Enhancing module to enhance the representation of high-resolution feature map.•Our fully-equipped model (EKENet), achieves the new state-of-the-art performance in terms of speed and accuracy tradeoff.

论文关键词：Scene parsing,Real-time method,Deep learning

论文评审过程：Received 7 October 2019, Revised 24 August 2020, Accepted 22 September 2020, Available online 25 September 2020, Version of Record 1 October 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107671