最新成果demo展示：

ECCV2022 | ChunkyGAN 又一个狠活

论文：https://dcgi.fel.cvut.cz/home/sykorad/Subrtova22-ECCV.pdf 代码：https://github.com/futscdav/Chunkmogrify

ECCV2022 汇总：https://github.com/DWCTOD/ECCV2022-Papers-with-Code-Demo/

摘要：

我们提出了 ChunkyGAN——一种使用生成对抗网络对图像进行建模和编辑的新范式。与先前寻求输入图像的全局潜在表示的技术不同，我们的方法将输入图像细分为一组较小的组件（块），这些组件可以手动或使用预训练的分割网络自动指定。对于每个块，由于约束数量较少，生成网络的潜在代码在本地以更高的准确度进行估计。此外，在潜码优化过程中，可以进一步细化分割以提高匹配质量。这个过程可以对原始图像进行高质量的投影，并实现以前的方法难以实现的空间解缠结。为了证明我们方法的优势，我们在各种图像编辑场景中对其进行了定量和定性评估，这些场景受益于该方法的更高重建质量和局部性质。我们的方法足够灵活，甚至可以操作使用全局技术难以重建的域外图像

最新论文整理

   ECCV2022

Updated on : 28 Jul 2022

total number : 16

ShAPO: Implicit Representations for Multi-Object Shape, Appearance, and Pose Optimization

• 论文/Paper: http://arxiv.org/pdf/2207.13691

• 代码/Code: None

Shift-tolerant Perceptual Similarity Metric

• 论文/Paper: http://arxiv.org/pdf/2207.13686

• 代码/Code: http://github.com/abhijay9/ShiftTolerant-LPIPS/

Abstracting Sketches through Simple Primitives

• 论文/Paper: http://arxiv.org/pdf/2207.13543

• 代码/Code: https://github.com/ExplainableML/sketch-primitives.

AutoTransition: Learning to Recommend Video Transition Effects

• 论文/Paper: http://arxiv.org/pdf/2207.13479

• 代码/Code: \url{https://github.com/acherstyx/AutoTransition}.

Hardly Perceptible Trojan Attack against Neural Networks with Bit Flips

• 论文/Paper: http://arxiv.org/pdf/2207.13417

• 代码/Code: https://github.com/jiawangbai/HPT

Identifying Hard Noise in Long-Tailed Sample Distribution

• 论文/Paper: http://arxiv.org/pdf/2207.13378

• 代码/Code: https://github.com/yxymessi/H2E-Framework

Efficient Video Deblurring Guided by Motion Magnitude

• 论文/Paper: http://arxiv.org/pdf/2207.13374

• 代码/Code: None

One-Trimap Video Matting

• 论文/Paper: http://arxiv.org/pdf/2207.13353

• 代码/Code: online:https://github.com/Hongje/OTVM.

PointFix: Learning to Fix Domain Bias for Robust Online Stereo Adaptation

• 论文/Paper: http://arxiv.org/pdf/2207.13340

• 代码/Code: None

SiRi: A Simple Selective Retraining Mechanism for Transformer-based Visual Grounding

• 论文/Paper: http://arxiv.org/pdf/2207.13325

• 代码/Code: None

Generator Knows What Discriminator Should Learn in Unconditional GANs

• 论文/Paper: http://arxiv.org/pdf/2207.13320

• 代码/Code: https://github.com/naver-ai/GGDR

End-to-end Graph-constrained Vectorized Floorplan Generation with Panoptic Refinement

• 论文/Paper: http://arxiv.org/pdf/2207.13268

• 代码/Code: None

Spatiotemporal Self-attention Modeling with Temporal Patch Shift for Action Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.13259

• 代码/Code: https://github.com/MartinXM/TPS.

Concurrent Subsidiary Supervision for Unsupervised Source-Free Domain Adaptation

• 论文/Paper: http://arxiv.org/pdf/2207.13247

• 代码/Code: None

Online Continual Learning with Contrastive Vision Transformer

• 论文/Paper: http://arxiv.org/pdf/2207.13516

• 代码/Code: None

LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity

• 论文/Paper: http://arxiv.org/pdf/2207.13129

• 代码/Code: None

   CVPR2022

Updated on : 28 Jul 2022

total number : 1

NICEST: Noisy Label Correction and Training for Robust Scene Graph Generation

• 论文/Paper: http://arxiv.org/pdf/2207.13316

• 代码/Code: None