论文/代码速递2022.10.13！

最新成果demo展示：

主页：https://sgvr.kaist.ac.kr/publication/flow-supervisor/ 代码：https://github.com/iwbn/flow-supervisor

 光流CNN的训练管道由合成数据集的预训练阶段和目标数据集的微调阶段组成。然而，从目标视频中获取ground truth 流需要付出巨大的努力。本文提出了一种实用的微调方法，以使预处理模型适应没有ground truth 流的目标数据集，这种方法尚未得到广泛的探索。具体来说，我们提出了一个用于自监督的流监督，它由参数分离和学生输出连接组成。这种设计的目的是稳定收敛，并比在微调任务中不稳定的传统自监督方法具有更好的精度。实验结果表明，与不同的自监督方法相比，该方法对于半监督学习是有效的。此外，通过利用额外的未标记数据集，我们在Sintel和KITTI基准上对最先进的光流模型进行了有意义的改进

最新论文整理

   ECCV2022

Updated on : 13 Oct 2022

total number : 1

DeepMend: Learning Occupancy Functions to Represent Shape for Repair

• 论文/Paper: http://arxiv.org/pdf/2210.05728

• 代码/Code: https://github.com/terascale-all-sensing-research-studio/deepmend

    CVPR2022

   NeurIPS

Updated on : 13 Oct 2022

total number : 11

AniFaceGAN: Animatable 3D-Aware Face Image Generation for Video Avatars

• 论文/Paper: http://arxiv.org/pdf/2210.06465

• 代码/Code: None

Learning Multi-resolution Functional Maps with Spectral Attention for Robust Shape Matching

• 论文/Paper: http://arxiv.org/pdf/2210.06373

• 代码/Code: https://github.com/craigleili/AttentiveFMaps

Latency-aware Spatial-wise Dynamic Networks

• 论文/Paper: http://arxiv.org/pdf/2210.06223

• 代码/Code: https://github.com/leaplabthu/lasnet

Multi-Granularity Cross-modal Alignment for Generalized Medical Visual Representation Learning

• 论文/Paper: http://arxiv.org/pdf/2210.06044

• 代码/Code: None

Long-Form Video-Language Pre-Training with Multimodal Temporal Contrastive Learning

• 论文/Paper: http://arxiv.org/pdf/2210.06031

• 代码/Code: https://github.com/microsoft/XPretrain.

Boosting the Transferability of Adversarial Attacks with Reverse Adversarial Perturbation

• 论文/Paper: http://arxiv.org/pdf/2210.05968

• 代码/Code: https://github.com/sclbd/transfer_attack_rap

Decomposed Knowledge Distillation for Class-Incremental Semantic Segmentation

• 论文/Paper: http://arxiv.org/pdf/2210.05941

• 代码/Code: None

A Lower Bound of Hash Codes' Performance

• 论文/Paper: http://arxiv.org/pdf/2210.05899

• 代码/Code: https://github.com/vl-group/lbhash

SegViT: Semantic Segmentation with Plain Vision Transformers

• 论文/Paper: http://arxiv.org/pdf/2210.05844

• 代码/Code: None

Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork

• 论文/Paper: http://arxiv.org/pdf/2210.06428

• 代码/Code: https://github.com/VITA-Group/Trap-and-Replace-Backdoor-Defense

Towards Theoretically Inspired Neural Initialization Optimization

• 论文/Paper: http://arxiv.org/pdf/2210.05956

• 代码/Code: https://github.com/HarborYuan/GradCosine