Timezone: »

SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation
Meng-Hao Guo · Cheng-Ze Lu · Qibin Hou · Zhengning Liu · Ming-Ming Cheng · Shi-min Hu

Tue Dec 06 05:00 PM -- 07:00 PM (PST) @

We present SegNeXt, a simple convolutional network architecture for semantic segmentation. Recent transformer-based models have dominated the field of se- mantic segmentation due to the efficiency of self-attention in encoding spatial information. In this paper, we show that convolutional attention is a more efficient and effective way to encode contextual information than the self-attention mech- anism in transformers. By re-examining the characteristics owned by successful segmentation models, we discover several key components leading to the perfor- mance improvement of segmentation models. This motivates us to design a novel convolutional attention network that uses cheap convolutional operations. Without bells and whistles, our SegNeXt significantly improves the performance of previous state-of-the-art methods on popular benchmarks, including ADE20K, Cityscapes, COCO-Stuff, Pascal VOC, Pascal Context, and iSAID. Notably, SegNeXt out- performs EfficientNet-L2 w/ NAS-FPN and achieves 90.6% mIoU on the Pascal VOC 2012 test leaderboard using only 1/10 parameters of it. On average, SegNeXt achieves about 2.0% mIoU improvements compared to the state-of-the-art methods on the ADE20K datasets with the same or fewer computations.

Author Information

Meng-Hao Guo (Tsinghua University)
Meng-Hao Guo

Third-year Ph.D candidate at the Department of Computer Science, Tsinghua University.

Cheng-Ze Lu
Qibin Hou (Nankai University)
Zhengning Liu (Tsinghua University, Tsinghua University)
Ming-Ming Cheng (Nankai University)
Shi-min Hu (Tsinghua University, Tsinghua University)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors