1 前言

本文盤點了CVPR 2022 目前為止的2D圖像分割相關(guān)論文，包含語義分割和實例分割，總計22篇論文，值得學(xué)習(xí)。

2 語義分割

2.1 強監(jiān)督

(1) ReSTR: Convolution-free Referring Image Segmentation Using Transformers

論文：https://arxiv.org/pdf/2203.16768.pdf

代碼：暫無

(2) Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation

論文：https://arxiv.org/pdf/2203.16768.pdf

代碼：https://github.com/jamycheung/Trans4PASS

(3) Deep Hierarchical Semantic Segmentation

論文：https://arxiv.org/pdf/2203.14335.pdf

代碼：https://github.com/0liliulei/HieraSeg

(4) Semantic Segmentation by Early Region Proxy

論文：https://arxiv.org/pdf/2203.14043.pdf

代碼：https://github.com/YiFZhang/RegionProxy

(5) SimT: Handling Open-set Noise for Domain Adaptive Semantic Segmentation

論文：https://arxiv.org/pdf/2203.15202.pdf

代碼：https://github.com/CityU-AIM-Group/SimT

(6) Rethinking Semantic Segmentation: A Prototype View

論文：https://arxiv.org/pdf/2203.15102.pdf

代碼：https://github.com/tfzhou/ProtoSeg

2.2 弱監(jiān)督

(1) Class Re-Activation Maps for Weakly-Supervised Semantic Segmentation

(弱監(jiān)督語義分割的類重新激活圖)

論文：https://arxiv.org/pdf/2203.00962.pdf

代碼：https://github.com/zhaozhengChen/ReCAM

(2) Multi-class Token Transformer for Weakly Supervised Semantic Segmentation

論文：https://arxiv.org/pdf/2203.02891.pdf

代碼：https://github.com/xulianuwa/MCTformer

(3) Learning Affinity from Attention: End-to-End Weakly-Supervised Semantic Segmentation with Transformers

論文：https://arxiv.org/pdf/2203.02664.pdf

代碼：https://github.com/rulixiang/afa

(4) Self-supervised Image-specific Prototype Exploration for Weakly Supervised Semantic Segmentation

論文：https://arxiv.org/pdf/2203.02909.pdf

代碼： https://github.com/chenqi1126/SIPE

(5) Cross Language Image Matching for Weakly Supervised Semantic Segmentation

論文：https://arxiv.org/pdf/2203.02668.pdf

代碼： https://github.com/CVISZU/CLIMS

(6) Weakly Supervised Semantic Segmentation using Out-of-Distribution Data

論文：https://arxiv.org/pdf/2203.03860.pdf

代碼：https://github.com/naver-ai/w-ood

(7) Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds

論文：https://arxiv.org/pdf/2203.16045.pdf

代碼：https://github.com/gaviotas/AMN

2.3 半監(jiān)督

(1) ST++: Make Self-training Work Better for Semi-supervised Semantic Segmentation

論文：https://arxiv.org/pdf/2106.05095.pdf

代碼：https://github.com/LiheYoung/ST-PlusPlus

(2) Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels

論文：https://arxiv.org/pdf/2203.03884.pdf

代碼：https://github.com/Haochen-Wang409/U2PL

2.4 無監(jiān)督

(1) GroupViT: Semantic Segmentation Emerges from Text Supervision

論文：https://arxiv.org/pdf/2202.11094.pdf

代碼： https://jerryxu.net/GroupViT

3 實例分割

3.1 強監(jiān)督

(1) BoxeR: Box-Attention for 2D and 3D Transformers

論文：https://arxiv.org/pdf/2111.13087.pdf

代碼：https://github.com/kienduynguyen/BoxeR.

(2) E2EC: An End-to-End Contour-based Method for High-Quality High-Speed Instance Segmentation

論文：https://arxiv.org/pdf/2203.04074.pdf

代碼：https://github.com/zhang-tao-whu/e2ec.

(3) Sparse Instance Activation for Real-Time Instance Segmentation

論文：https://arxiv.org/pdf/2203.12827.pdf

代碼：https://github.com/hustvl/SparseInst

(4) SharpContour: A Contour-based Boundary Refinement Approach for Efficient and Accurate Instance Segmentation

論文：https://arxiv.org/pdf/2203.13312.pdf

代碼：暫無

3.2 半監(jiān)督

(1) Noisy Boundaries: Lemon or Lemonade for Semi-supervised Instance Segmentation?

論文：https://arxiv.org/pdf/2203.13427.pdf

代碼：https://github.com/zhenyuw16/noisyboundaries

3.3 無監(jiān)督

(1) FreeSOLO: Learning to Segment Objects without Annotations

論文：https://arxiv.org/pdf/2202.12181.pdf

代碼：暫無

文章來源地址http://www.zghlxwxcb.cn/news/detail-518292.html

到了這里，關(guān)于CVPR 2022: 圖像分割論文大盤點的文章就介紹完了。如果您還想了解更多內(nèi)容，請在右上角搜索TOY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！