当前位置:网站首页>浅谈CVPR2022的几个研究热点
浅谈CVPR2022的几个研究热点
2022-07-01 07:18:00 【程序猿老甘】
CVPR2022刚刚结束,作为影响力最广的视觉盛会,今年又有一批优秀的工作被展示出来。相信关注视觉最新研究进展的各位小伙伴,已经磨拳擦掌,准备向CVPR2023投稿了。基于今年的工作,到底哪些领域是CVPR关注的热点?哪些领域的工作,接受度更高,oral的比例更大呢?基于CVPR官方最新的统计信息,我将跟大家聊聊CVPR的一些研究热点,希望对那些计划投下一轮CVPR的同学提供一点参考信息。
1. 十大热点研究领域
首先,我们基于oral论文的统计信息,按照接收论文比重以及所述领域进行排序,得到的十个热点领域,包括:多角度三维视觉,图像与视频合成,识别检测分类与检索,深度网络结构设计,视觉与语言处理交叉,低质量数据视觉分析,形状分析,迁移学习,视频分析与理解,姿态估计。

图1. 十大研究热点领域(Oral)
当我们统计全部接收论文时,统计数据在顺序上会有一点变化,包括:识别检测分类与检索,图像与视频合成,多角度三维视觉,低质量数据视觉分析,视觉与语言处理交叉,形状分析,迁移学习,深度网络结构设计,自监督与非监督学习,视频分析与理解。

图2. 十大研究热点领域(All)
可以看到,两个排序对应的研究热点问题,具有极高的重复性。结合两个表,偏重于应用层面的角度对热点进行总结,我从中选出五个热点研究方向,供计划投稿的同学参考:
- 多角度三维视觉
- 图像与视频合成
- 识别检测分类与检索
- 视觉与语言处理交叉
- 低质量数据视觉分析
2. Best Paper
CVPR2022的Best paper list包含四篇文章,分别为:
Best Paper Award: Learning to Solve Hard Minimal Problems
Best Paper Honorable Mention: Dual-Shutter Optical Vibration Sensing
Best Student Paper Award: EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation
Best Student Paper Honorable Mention: Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields
最佳论文为《Learning to Solve Hard Minimal Problems》。粗看了下,不是很懂,大概是在对优化问题领域做了一些偏理论性的工作,引入了几何优化的一些工具。《Dual-Shutter Optical Vibration Sensing》是关于三维激光扫描的技术。《EPro-PnP: Generalized End-to-End Probabilistic Perspective...》基于多点透视理论,提出一种从图像中估计物体的三维姿态的方法。《Ref-NeRF》基本就是NeRF算法的变种研究。从最佳论文的侧重可以知道,CVPR比较青睐三维视觉相关研究。另外,会前呼声较高的Kaiming老师的《Masked Autoencoders Are Scalable Vision Learners》也是值得深入学习的。基于MAE提出一种基于patch预测的编解码结构,对于数据图像内容理解具有极好的预测与重建性能。该论文被列为最佳论文候选。
3. 个人关注
因为我个人最近一直在做颜色迁移,光照优化一类的工作,所以比较关注low-level vision领域。今年CVPR在该领域录取了19篇oral以及91篇poster,接收文章数不能算少。我将对应的19篇oral文章抄写在这里,方便之后学习。
[1] Robust Equivariant Imaging: A Fully Unsupervised Framework for Learning To Image From Noisy and Partial Measurements. (去噪+超分辨率用于图像增强技术)
[2] Bijective Mapping Network for Shadow Removal. (消除影子)
[3] Event-Aided Direct Sparse Odometry. (稀疏点云加强)
[4] MAXIM: Multi-Axis MLP for Image Processing.(通用图像质量增强算法)
[5] Details or Artifacts: A Locally Discriminative Learning Approach to Realistic Image Super-Resolution.(超分辨率)
[6] Dual Adversarial Adaptation for Cross-Device Real-World Image Super-Resolution. (超分辨率)
[7] ELIC: Efficient Learned Image Compression With Unevenly Grouped Space-Channel Contextual Adaptive Coding.
[8] Discrete Cosine Transform Network for Guided Depth Map Super-Resolution. (超分辨率)
[9] Deep Rectangling for Image Stitching: A Learning Baseline.(图像拼接)
[10] CamLiFlow: Bidirectional Camera-LiDAR Fusion for Joint Optical Flow and Scene Flow Estimation. (光流优化)
[11] Toward Fast, Flexible, and Robust Low-Light Image Enhancement. (低光增强)
[12] Faithful Extreme Rescaling via Generative Prior Reciprocated Invertible Represe-ntations.
[13] Learning Trajectory-Aware Transformer for Video Super-Resolution. (超分辨率)
[14] SphereSR: 360deg Image Super-Resolution With Arbitrary Projection via Continuous Spherical Image Representation.(超分辨率)
[15] Parametric Scattering Networks. (优化的学习结构)
[16] Target-Aware Dual Adversarial Learning and a Multi-Scenario Multi-Modality Benchmark To Fuse Infrared and Visible for Object Detection. (低光环境下的对象探测)
[17] Learning to Deblur Using Light Field Generated and Real Defocus Images. (去模糊)
[18] Burst Image Restoration and Enhancement. (图像重建)
[19 ]Restormer: Efficient Transformer for High-Resolution Image Restoration. (去模糊)
在low-level vision领域,超分辨率仍然占有较大的比重。一些工作包括去模糊,质量增强,细节重建等,本质上还是和超分辨率技术有紧密的联系。看来,未来做low-level vision,大概率要利用到超分辨率算法。从部分论文可以看出,三维视觉已经结合到low-level vision领域。针对深度图,全景照片等具有三维属性的数据,进行细节重建,运动补偿等计算,也是很不错的研究方向。
边栏推荐
- [Electrical dielectric number] electrical dielectric number and calculation considering HVDC and facts components
- [recommendation technology] matlab simulation of network information recommendation technology based on collaborative filtering
- C language implementation [minesweeping game] full version (implementation source code)
- 5G Massive MIMO的概念和优点总结
- weback5基础配置详解
- Are there any practical skills for operation and maintenance management
- How the esp32 deep sleep current is lower than 10uA
- 手机开户选哪个证券公司比较好,哪个更安全
- 【计网】(一) 集线器、网桥、交换机、路由器等概念
- Understanding of Turing test and Chinese Room
猜你喜欢
![[recommendation technology] matlab simulation of network information recommendation technology based on collaborative filtering](/img/fb/dc03f97f12488e53d706a05da9faea.png)
[recommendation technology] matlab simulation of network information recommendation technology based on collaborative filtering

【计网】(一) 集线器、网桥、交换机、路由器等概念

比赛即实战!中国软件杯发布全新产业创新赛项,校企可联合参赛

Alibaba OSS postman invalid according to policy: policy condition failed: ["starts with", "key", "test/"]

C # read and write customized config file

ctfshow-web352,353(SSRF)

北漂程序员深夜emo发帖求助:女朋友走了我很孤独 ......
![C language implementation [Sanzi chess game] (step analysis and implementation source code)](/img/3b/d32b46292ed20f31a6e1db97349df1.png)
C language implementation [Sanzi chess game] (step analysis and implementation source code)
![[programming training 2] sorting subsequence + inverted string](/img/96/87750c5d3954ef6c39cce073e8b9ae.png)
[programming training 2] sorting subsequence + inverted string

未来互联网人才还稀缺吗?哪些技术方向热门?
随机推荐
[target detection] yolov5, the shoulder of target detection (detailed principle + Training Guide)
Mysql与Redis一致性解决方案
ctfshow-web351(SSRF)
【编程强训2】排序子序列+倒置字符串
Are there any practical skills for operation and maintenance management
【推荐技术】基于协同过滤的网络信息推荐技术matlab仿真
Is it reliable to open an account on the compass with your mobile phone? Is there any potential safety hazard
华泰证券开户是安全可靠的么?怎么开华泰证券账户
Image style migration cyclegan principle
Vscode automatically formats code according to eslint specification
[FPGA frame difference] FPGA implementation of frame difference target tracking based on vmodcam camera
开源了!文心大模型ERNIE-Tiny轻量化技术,又准又快,效果全开
[Tikhonov] image super-resolution reconstruction based on Tikhonov regularization
The computer has a network, but all browser pages can't be opened. What's the matter?
【微服务|openfeign】Feign的日志记录
【深圳IO】精确食品称(汇编语言的一些理解)
解决kaniko push镜像到harbor时报错(代理导致):unexpected status code 503 Service Unavailable
灰度何以跌下神坛?
Fix the problem that the AI video intelligent platform easycvr device video cannot be played
[matlab] solve nonlinear programming