ECCV2022 &CVPR2022论文速递2022.7.22！47篇 demo

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：

ECCV2022 | XMem: 高质量长期视频分割！

效果超群！

标题：XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model

论文：https://arxiv.org/pdf/2207.07115.pdf

代码：https://github.com/hkchengrex/XMem

摘要：

我们提出了 XMem，这是一种用于长视频的视频对象分割架构，具有统一的特征内存存储，受 Atkinson-Shiffrin 内存模型的启发。先前关于视频对象分割的工作通常只使用一种类型的特征记忆。对于超过一分钟的视频，单个特征内存模型将内存消耗和准确性紧密联系在一起。相比之下，遵循 Atkinson-Shiffrin 模型，我们开发了一种架构，该架构包含多个独立但深度连接的特征记忆存储：快速更新的感觉记忆、高分辨率工作记忆和紧凑的持续长期记忆。至关重要的是，我们开发了一种记忆增强算法，该算法通常将积极使用的工作记忆元素整合到长期记忆中，从而避免记忆爆炸并最大限度地减少长期预测的性能衰减。结合新的内存读取机制，XMem 在长视频数据集上的性能大大超过了最先进的性能，同时在短视频上与最先进的方法（不适用于长视频）相当数据集。

最新论文整理

ECCV2022

Updated on : 22 Jul 2022

total number : 47

Online Domain Adaptation for Semantic Segmentation in Ever-Changing Conditions

• 论文/Paper: http://arxiv.org/pdf/2207.10667
• 代码/Code: https://github.com/theo2021/onda

TinyViT: Fast Pretraining Distillation for Small Vision Transformers

• 论文/Paper: http://arxiv.org/pdf/2207.10666
• 代码/Code: https://github.com/microsoft/cream

Exploring Fine-Grained Audiovisual Categorization with the SSW60 Dataset

• 论文/Paper: http://arxiv.org/pdf/2207.10664
• 代码/Code: https://github.com/visipedia/ssw60

In Defense of Online Models for Video Instance Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.10661
• 代码/Code: https://github.com/wjf5203/vnext

Novel Class Discovery without Forgetting

• 论文/Paper: http://arxiv.org/pdf/2207.10659
• 代码/Code: None

Generative Multiplane Images: Making a 2D GAN 3D-Aware

• 论文/Paper: http://arxiv.org/pdf/2207.10642
• 代码/Code: https://github.com/apple/ml-gmpi

Approximate Differentiable Rendering with Algebraic Surfaces

• 论文/Paper: http://arxiv.org/pdf/2207.10606
• 代码/Code: None

Unsupervised Night Image Enhancement: When Layer Decomposition Meets Light-Effects Suppression

• 论文/Paper: http://arxiv.org/pdf/2207.10564
• 代码/Code: https://github.com/jinyeying/night-enhancement

An Efficient Spatio-Temporal Pyramid Transformer for Action Detection

• 论文/Paper: http://arxiv.org/pdf/2207.10448
• 代码/Code: None

Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration

• 论文/Paper: http://arxiv.org/pdf/2207.10447
• 代码/Code: https://github.com/164140757/scm

Mining Relations among Cross-Frame Affinities for Video Semantic Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.10436
• 代码/Code: https://github.com/guoleisun/vss-mrcfa

Human Trajectory Prediction via Neural Social Physics

• 论文/Paper: http://arxiv.org/pdf/2207.10435
• 代码/Code: https://github.com/realcrane/human-trajectory-prediction-via-neural-social-physics

D2-TPred: Discontinuous Dependency for Trajectory Prediction under Traffic Lights

• 论文/Paper: http://arxiv.org/pdf/2207.10398
• 代码/Code: https://github.com/vtp-tl/d2-tpred

FADE: Fusing the Assets of Decoder and Encoder for Task-Agnostic Upsampling

• 论文/Paper: http://arxiv.org/pdf/2207.10392
• 代码/Code: None

Error Compensation Framework for Flow-Guided Video Inpainting

• 论文/Paper: http://arxiv.org/pdf/2207.10391
• 代码/Code: None

NSNet: Non-saliency Suppression Sampler for Efficient Video Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.10388
• 代码/Code: None

Pose for Everything: Towards Category-Agnostic Pose Estimation

• 论文/Paper: http://arxiv.org/pdf/2207.10387
• 代码/Code: https://github.com/luminxu/Pose-for-Everything.

Temporal Saliency Query Network for Efficient Video Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.10379
• 代码/Code: None

LocVTP: Video-Text Pre-training for Temporal Localization

• 论文/Paper: http://arxiv.org/pdf/2207.10362
• 代码/Code: https://github.com/mengcaopku/locvtp

CADyQ: Content-Aware Dynamic Quantization for Image Super-Resolution

• 论文/Paper: http://arxiv.org/pdf/2207.10345
• 代码/Code: https://github.com/cheeun/cadyq

UFO: Unified Feature Optimization

• 论文/Paper: http://arxiv.org/pdf/2207.10341
• 代码/Code: None

OIMNet++: Prototypical Normalization and Localization-aware Learning for Person Search

• 论文/Paper: http://arxiv.org/pdf/2207.10320
• 代码/Code: None

AutoAlignV2: Deformable Feature Aggregation for Dynamic Multi-Modal 3D Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.10316
• 代码/Code: https://github.com/zehuichen123/autoalignv2

SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformer

• 论文/Paper: http://arxiv.org/pdf/2207.10315
• 代码/Code: https://github.com/hrzhou2/seedformer

AdaNeRF: Adaptive Sampling for Real-time Rendering of Neural Radiance Fields

• 论文/Paper: http://arxiv.org/pdf/2207.10312
• 代码/Code: None

Towards Accurate Open-Set Recognition via Background-Class Regularization

• 论文/Paper: http://arxiv.org/pdf/2207.10287
• 代码/Code: None

Grounding Visual Representations with Texts for Domain Generalization

• 论文/Paper: http://arxiv.org/pdf/2207.10285
• 代码/Code: https://github.com/mswzeus/gvrt

DeltaGAN: Towards Diverse Few-shot Image Generation with Sample-Specific Delta

• 论文/Paper: http://arxiv.org/pdf/2207.10271
• 代码/Code: https://github.com/bcmi/deltagan-few-shot-image-generation

Injecting 3D Perception of Controllable NeRF-GAN into StyleGAN for Editable Portrait Image Synthesis

• 论文/Paper: http://arxiv.org/pdf/2207.10257
• 代码/Code: https://github.com/jgkwak95/surf-gan

SGBANet: Semantic GAN and Balanced Attention Network for Arbitrarily Oriented Scene Text Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.10256
• 代码/Code: None

SPIN: An Empirical Evaluation on Sharing Parameters of Isotropic Networks

• 论文/Paper: http://arxiv.org/pdf/2207.10237
• 代码/Code: https://github.com/apple/ml-spin

MeshMAE: Masked Autoencoders for 3D Mesh Data Analysis

• 论文/Paper: http://arxiv.org/pdf/2207.10228
• 代码/Code: None

On Label Granularity and Object Localization

• 论文/Paper: http://arxiv.org/pdf/2207.10225
• 代码/Code: https://github.com/visipedia/inat_loc

Spotting Temporally Precise, Fine-Grained Events in Video

• 论文/Paper: http://arxiv.org/pdf/2207.10213
• 代码/Code: None

2D GANs Meet Unsupervised Single-view 3D Reconstruction

• 论文/Paper: http://arxiv.org/pdf/2207.10183
• 代码/Code: None

Controllable and Guided Face Synthesis for Unconstrained Face Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.10180
• 代码/Code: None

Video Anomaly Detection by Solving Decoupled Spatio-Temporal Jigsaw Puzzles

• 论文/Paper: http://arxiv.org/pdf/2207.10172
• 代码/Code: None

GOCA: Guided Online Cluster Assignment for Self-Supervised Video Representation Learning

• 论文/Paper: http://arxiv.org/pdf/2207.10158
• 代码/Code: https://github.com/seleucia/goca

Visual Knowledge Tracing

• 论文/Paper: http://arxiv.org/pdf/2207.10157
• 代码/Code: https://github.com/nkondapa/visualknowledgetracing

Tackling Long-Tailed Category Distribution Under Domain Shifts

• 论文/Paper: http://arxiv.org/pdf/2207.10150
• 代码/Code: https://github.com/guxiao0822/lt-ds

Latent Discriminant deterministic Uncertainty

• 论文/Paper: http://arxiv.org/pdf/2207.10130
• 代码/Code: https://github.com/ensta-u2is/ldu

Animation from Blur: Multi-modal Blur Decomposition with Motion Guidance

• 论文/Paper: http://arxiv.org/pdf/2207.10123
• 代码/Code: https://github.com/zzh-tech/Animation-from-Blur.

BRACE: The Breakdancing Competition Dataset for Dance Motion Synthesis

• 论文/Paper: http://arxiv.org/pdf/2207.10120
• 代码/Code: https://github.com/dmoltisanti/brace

Bitwidth-Adaptive Quantization-Aware Neural Network Training: A Meta-Learning Approach

• 论文/Paper: http://arxiv.org/pdf/2207.10188
• 代码/Code: None

Structural Causal 3D Reconstruction

• 论文/Paper: http://arxiv.org/pdf/2207.10156
• 代码/Code: None

AudioScopeV2: Audio-Visual Attention Architectures for Calibrated Open-Domain On-Screen Sound Separation

• 论文/Paper: http://arxiv.org/pdf/2207.10141
• 代码/Code: None

Continual Variational Autoencoder Learning via Online Cooperative Memorization

• 论文/Paper: http://arxiv.org/pdf/2207.10131
• 代码/Code: https://github.com/dtuzi123/ovae

CVPR2022