VALSE 论文速览 第79期：从原型角度重新审视语义分割

论文题目：Rethinking Semantic Segmentation: A Prototype View

作者列表：Tianfei Zhou (ETH Zurich)，Wenguan Wang (University of Technology Sydney)，Ender Konukoglu (ETH Zurich), Luc Van Gool (ETH Zurich)

B站观看网址：

https://www.bilibili.com/video/BV1qW4y167CK/

复制链接到浏览器打开或点击阅读原文即可跳转至观看页面。

论文摘要：

Prevalent semantic segmentation solutions, despite their different network designs (FCN based or attention based)and mask decoding strategies (parametric softmax based or pixel-query based), can be placed in one category, by considering the softmax weights or query vectors as learnable class prototypes. In light of this prototype view, this study uncovers several limitations of such parametric segmentation regime, and proposes a nonparametric alternative based on non-learnable prototypes. Instead of prior methods learning a single weight/query vector for each class in a fully parametric manner, our model represents each class as a set of non-learnable prototypes, relying solely on the mean features of several training pixels within that class. The dense prediction is thus achieved by nonparametric nearest prototype retrieving. This allows our model to directly shape the pixel embedding space, by optimizing the arrangement between embedded pixels and anchored prototypes. It is able to handle arbitrary number of classes with a constant amount of learnable parameters. We empirically show that, with FCN based and attention based segmentation models (i.e., HRNet, Swin, SegFormer)and backbones (i.e., ResNet, HRNet, Swin, MiT), our nonparametric framework yields compelling results over several datasets (i.e., ADE20K, Cityscapes, COCO-Stuff), and performs well in the large-vocabulary situation. We expect this work will provoke a rethink of the current de facto semantic segmentation model design.

论文信息：

[1] Tianfei Zhou, Wenguan Wang, Ender Konukoglu and Luc Van Gool. Rethinking Semantic Segmentation: A Prototype View. CVPR 2022 (Oral)

论文链接：

[https://arxiv.org/pdf/2203.15102.pdf]

代码链接：

[https://github.com/tfzhou/ProtoSeg]

视频讲者简介：

周天飞，苏黎世联邦理工学院CVL博后研究员，主要研究方向是计算机视觉和深度学习。

特别鸣谢本次论文速览主要组织者：

月度轮值AC：赵文达 (大连理工大学)、任文琦 (中山大学)

季度责任AC：魏秀参 (南京理工大学)