论文/代码速递2022.11.1！

整理：AI算法与图像处理

CVPR2022论文和代码整理：https://github.com/DWCTOD/CVPR2022-Papers-with-Code-Demo

ECCV2022论文和代码整理：https://github.com/DWCTOD/ECCV2022-Papers-with-Code-Demo

最新成果demo展示：

标题：MuLUT: Cooperating Mulitple Look-Up Tables for Efficient Image Super-Resolution

论文：

https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780234.pdf

主页：https://mulut.pages.dev/

代码：https://github.com/ddlee-cn/MuLUT

边缘设备的高分辨率屏幕刺激了对高效图像超分辨率（SR）的强烈需求。一项新兴的研究，SR-LUT，通过将查找表（LUT）与基于学习的SR方法相结合来响应这一需求。然而，单个LUT的大小随着其索引容量的增加呈指数级增长。因此，单个LUT的感受野受到限制，导致性能低下。为了解决这个问题，我们通过支持多个LUT（称为MuLUT）的合作来扩展SR-LUT。首先，我们设计了两种新的互补索引模式，并行构造了多个LUT。其次，我们提出了一种重新索引机制，以实现多个LUT之间的分层索引。在这两种方式中，MuLUT的总大小与其索引容量呈线性关系，从而产生了一种获得优异性能的实用方法。我们检查了MuLUT在五个SR基准上的优势。MuLUT比SR-LUT实现了显著的改进，最高可达1.1dB PSNR，同时保持了其效率。此外，我们扩展了MuLUT以解决拜耳图案图像的去马赛克问题，在两个基准上大大超过了SR-LUT。

最新论文整理

ECCV2022

Updated on : 1 Nov 2022

total number : 3

Max Pooling with Vision Transformers reconciles class and shape in weakly supervised semantic segmentation

• 论文/Paper: http://arxiv.org/pdf/2210.17400
• 代码/Code: https://github.com/deepplants/vit-pcm

Generative Negative Text Replay for Continual Vision-Language Pretraining

• 论文/Paper: http://arxiv.org/pdf/2210.17322
• 代码/Code: None

Time-rEversed diffusioN tEnsor Transformer: A new TENET of Few-Shot Object Detection

• 论文/Paper: http://arxiv.org/pdf/2210.16897
• 代码/Code: https://github.com/zs123-lang/tenet

CVPR2022

NeurIPS

Updated on : 1 Nov 2022

total number : 10

Deep Model Reassembly

• 论文/Paper: http://arxiv.org/pdf/2210.17409
• 代码/Code: https://github.com/Adamdad/DeRy

Unified Optimal Transport Framework for Universal Domain Adaptation

• 论文/Paper: http://arxiv.org/pdf/2210.17067
• 代码/Code: https://github.com/changwxx/uniot-for-unida

Alleviating the Sample Selection Bias in Few-shot Learning by Removing Projection to the Centroid

• 论文/Paper: http://arxiv.org/pdf/2210.16834
• 代码/Code: https://github.com/kikimormay/fsl-tcbr

Towards Versatile Embodied Navigation

• 论文/Paper: http://arxiv.org/pdf/2210.16822
• 代码/Code: https://github.com/hanqingwangai/VXN

SL3D: Self-supervised-Self-labeled 3D Recognition

• 论文/Paper: http://arxiv.org/pdf/2210.16810
• 代码/Code: https://github.com/fcendra/sl3d

Dataset Distillation via Factorization

• 论文/Paper: http://arxiv.org/pdf/2210.16774
• 代码/Code: https://github.com/huage001/datasetfactorization

Breaking the Symmetry: Resolving Symmetry Ambiguities in Equivariant Neural Networks

• 论文/Paper: http://arxiv.org/pdf/2210.16646
• 代码/Code: None

Learning Probabilistic Models from Generator Latent Spaces with Hat EBM

• 论文/Paper: http://arxiv.org/pdf/2210.16486
• 代码/Code: https://github.com/point0bar1/hat-ebm

Region of Interest Detection in Melanocytic Skin Tumor Whole Slide Images

• 论文/Paper: http://arxiv.org/pdf/2210.16457
• 代码/Code: https://github.com/cyMichael/ROI_Detection

Learning Audio-Visual Dynamics Using Scene Graphs for Audio Source Separation

• 论文/Paper: http://arxiv.org/pdf/2210.16472
• 代码/Code: None