FaPN: Feature-aligned Pyramid Network for Dense Image Prediction

Huang，Shihua; Lu，Zhichao; Cheng，Ran; He，Cheng

10.1109/ICCV48922.2021.00090

2021

1550-5499

978-1-6654-2813-2

Proceedings of the IEEE International Conference on Computer Vision

844-853

10-17 Oct. 2021

Montreal, QC, Canada

Recent advancements in deep neural networks have made remarkable leap-forwards in dense image prediction. However, the issue of feature alignment remains as neglected by most existing approaches for simplicity. Direct pixel addition between upsampled and local features leads to feature maps with misaligned contexts that, in turn, translate to mis-classifications in prediction, especially on object boundaries. In this paper, we propose a feature alignment module that learns transformation offsets of pixels to contextually align upsampled higher-level features; and another feature selection module to emphasize the lower-level features with rich spatial details. We then integrate these two modules in a top-down pyramidal architecture and present the Feature-aligned Pyramid Network (FaPN). Extensive experimental evaluations on four dense prediction tasks and four datasets have demonstrated the efficacy of FaPN, yielding an overall improvement of 1.2 - 2.6 points in AP/mIoU over FPN when paired with Faster/Mask R-CNN. In particular, our FaPN achieves the state-of-the-art of 56.7% mIoU on ADE20K when integrated within MaskFormer. The code is available from https://github.com/EMIGroup/FaPN.

Recognition and classification Detection and localization in 2D and 3D grouping and shape Segmentation

第一

英语

EI

National Natural Science Foundation of China[61903178];National Natural Science Foundation of China[61906081];National Natural Science Foundation of China[U20A20306];

20221511951285

2-s2.0-85121798784

Scopus

Department of Computer Science and Engineering,Southern University of Science and Technology,

Huang，Shihua,Lu，Zhichao,Cheng，Ran,et al. FaPN: Feature-aligned Pyramid Network for Dense Image Prediction[C],2021:844-853.