ChengYue's Cat

✈️ 2022-05

🐯 2022-05-27

🍉 arXiv Paper

• Fast Vision Transformers with HiLo Attention
  关键词：Transformer, Attention
• MixMIM: Mixed and Masked Image Modeling for Efficient Visual Representation Learning
  关键词：自监督学习, Masked Image Modeling
• HIRL: A General Framework for Hierarchical Image Representation Learning
  关键词： Hierarchical 层级语义表征学习
• SemAffiNet: Semantic-Affine Transformation for Point Cloud Segmentation
  关键词：3D点云分割, 语义仿射变换
• AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognition
  关键词：Transformer
• Revealing the Dark Secrets of Masked Image Modeling
  关键词： 解释 Masked Image Modeling

🍉 工作内容与进展

• 旷视实习

🍉 感触

• 当前改进Vision Transformer的工作热度还是很高。

🌙 🌙 🌙

---

🐯 2022-05-28

---

🐯 2022-05-29

---

🐯 2022-06-01

🍉 arXiv Paper

• GCoNet+: A Stronger Group Collaborative Co-Salient Object Detector
  关键词：Co-Salient 目标检测

• What Knowledge Gets Distilled in Knowledge Distillation?
  关键词：知识蒸馏

• Self-Supervised Visual Representation Learning with Semantic Grouping
  关键词：自监督学习 语义分组
  感觉自己对自监督学习这块的工作有点脸盲，不是很好地鉴别其创新的地方到底在哪？？

• Rethinking Spatial Dimensions of Vision Transformers
  关键词：Transformer
  其实是2021-ICCV的工作了，这里也把挂出来方便以后学习查阅。

• FP-DETR: Detection Transformer Advanced by Fully Pre-training
  关键词: Transformer, Detection, 预训练

🍉 代码开源

• CCAM: Contrastive learning of Class-agnostic Activation Map for Weakly Supervised Object Localization and Semantic Segmentation 代码开源

• CLIMS: Cross Language Image Matching for Weakly Supervised Semantic Segmentation 代码开源

🍉 工作内容与进展

• 旷视实习这边的工作进展不高，很丧…
• 简单跑了下 CCAM源码， 很简单的工作，work 程度也一般吧，不过idea比较新颖。
• 最近需要准备下IJCAI2022的 presentation，以及中期答辩材料。

🍉 感触

• 自律自强

---

🐯 2022-06-03

🍉 arXiv Paper

• Siamese Image Modeling for Self-Supervised Vision Representation Learning
  关键词：自监督学习

• Semantic Instance Segmentation of 3D Scenes Through Weak Bounding Box Supervision
  关键词：3D弱监督学习

• Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features
  关键词：

• Optimizing Relevance Maps of Vision Transformers Improves Robustness
  关键词： Transformers

• VL-BEIT: Generative Vision-Language Pretraining
  关键词：视觉-语言预训练

• Prefix Conditioning Unifies Language and Label Supervision
  关键词：预训练

• Where are my Neighbors? Exploiting Patches Relations in Self-Supervised Vision Transformer
  关键词：Transformer

🍉 工作内容与进展

🍉 感触

• 都在大规模预训练。。。。这玩意到底能走多远？有多少潜力呢？
• 卷起来

🍉 计划与安排

• 阅读 Domain Adaptive Semantic Segmentation With Self-Supervised Depth Estimation
• 学习 timm 的模型构造函数

---

🐯 2022-06-09

🍉 arXiv Paper

• NeurIPS_2022_在投 Towards Understanding Why Mask-Reconstruction Pretraining Helps in Downstream Tasks

🍉 工作内容与进展

• 醉了，深度估计工作价值不高，语义相关性嘛，还在努力探索。

🍉 计划与安排

• 了解一些语义相关性特别是弱监督语义相关性的工作

• CVPR_2018 End-to-end weakly-supervised semantic alignment
• ECCV_2020 DHPF: Learning to Compose Hypercolumns for Visual Correspondence
• ICCV_2021 DISCOBOX: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

---

🐯 2022-06-14

🍉 arXiv Paper

• NeurIPS_2022在投Discovering Object Masks with Transformers for Unsupervised Semantic Segmentation
• NeurIPS_2022在投EnergyMatch: Energy-based Pseudo-Labeling for Semi-Supervised Learning
• ICML_2022_Learning Domain Adaptive Object Detection with Probabilistic Teacher
• NeurIPS_2022在投_Featurized Query R-CNN
• NeurIPS_2022在投_Better Teacher Better Student: Dynamic Prior Knowledge for Knowledge Distillation
• NeurIPS_2022在投_GLIPv2: Unifying Localization and VL Understanding

🍉 工作内容与进展

• 把IJCAI poster做一下。

---

🐯 2022-06-16

🍉 arXiv Paper

• Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone
• SP-ViT: Learning 2D Spatial Priors for Vision Transformers
• Visual Parser: Representing Part-whole Hierarchies with Transformers
• Masked Siamese ConvNets
• Masked Frequency Modeling for Self-Supervised Visual Pre-Training
• Grounded Language-Image Pre-training

---

🐯 2022-06-20

🍉 arXiv Paper

• NeurIPS2022_Under_Review Rectify ViT Shortcut Learning by Visual Saliency
• ICML_2022_VLMixer: Unpaired Vision-Language Pre-training via Cross-Modal CutMix
• ECCV_2022_在投_Learning Implicit Feature Alignment Function for Semantic Segmentation
• ECCV_2022_在投_Cross-task Attention Mechanism for Dense Multi-task Learning
• BatchFormer: Learning to Explore Sample Relationships for Robust Representation Learning 开源代码
• BatchFormerV2:Exploring Sample Relationships for Dense Representation Learning

---

🐯 2022-06-21

🍉 arXiv Paper

今天周二唉，竟然没有更新paper， 不明所以。 看牛客一哥们说的自监督/多模态预训练和扩散模型接下来会是接替Transformer和GAN的热点。

• 由浅入深了解Diffusion Model
• simplest implement of Denoising Diffusion Probability Model
• 大佬的博客讲解
• Awesome-Diffusion-Models

🍉 找工作进展

• 智能驾驶 华为车bu智能驾驶部，听知乎说加班比较严重。

• 大厂研究院

  • 华为诺亚-知乎实习
  • 阿里达摩院多模态-牛客实习貌似已经截止
  • 阿里达摩院-灵瞳实验室
  • 阿里达摩院-城市大脑
  • 腾讯优图

感觉华为诺亚可以搞一波？

🐯 2022-06-22

🍉 arXiv Paper

• Vicinity Vision Transformer
• Global Context Vision Transformers
• Bi-Calibration Networks for Weakly-Supervised Video Representation Learning
• SemMAE: Semantic-Guided Masking for Learning Masked Autoencoders
• KE-RCNN: Unifying Knowledge based Reasoning into Part-level Attribute Parsing
• Saliency Guided Inter- and Intra-Class Relation Constraints for Weakly Supervised Semantic Segmentation
• Unbiased Teacher v2: Semi-supervised Object Detection for Anchor-free and Anchor-based Detectors
• What is Where by Looking: Weakly-Supervised Open-World Phrase-Grounding without Text Inputs
• DALL-E for Detection: Language-driven Context Image Synthesis for Object Detection
• AAAI2021_Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization
• VISUALIZING AND UNDERSTANDING SELF-SUPERVISED VISION LEARNING
• Bear the Query in Mind: Visual Grounding with Query-conditioned Convolution
• NeurIPS_2022_Intra-Instance VICReg: Bag of Self-Supervised Image Patch Embedding
• CVPR_2022_oral_CMT-DeepLab: Clustering Mask Transformers for Panoptic Segmentation
• NeurIPS_2022_A Flexible Diffusion Model

---

🐯 2022-06-23

🍉 arXiv Paper

• Weakly-supervised Action Localization via Hierarchical Mining
• TiCo: Transformation Invariance and Covariance Contrast for Self-Supervised Visual Representation Learning
• SCIM: Simultaneous Clustering, Inference, and Mapping for Open-World Semantic Scene Understanding
• No Attention is Needed: Grouped Spatial-temporal Shift for Simple and Efficient Video Restorers
• Fighting Fire with Fire: Avoiding DNN Shortcuts through Priming
• Learning Debiased Classifier with Biased Committee
• Multi-level Domain Adaptation for Lane Detection
• Open Vocabulary Object Detection with Proposal Mining and Prediction Equalization

---

🐯 2022-06-24

🍉 arXiv Paper

TODO

🍉 工作内容与进展

起实验：

• 完善评测代码后，验证：
  • 在使用数据增强 (RS-RC-CJ) 后的 Baseline
    • TS-CAM-(RS-RC-CJ)-(BatchSize-64-2) ☑
    • TS-CAM-(RS-RC-CJ)-(BatchSize-100-2) ☑
  • (CCAM-Neg-Last2BLK-1.0T)的作用
    • GPTR-(CCAM-Neg-Last2BLK-1.0T)-(RS-RC-CJ)-(BatchSize-50-2) ☑
    • GPTR-(CCAM-Neg-Last2BLK-1.0T)-(RS-RC-CJ)-(BatchSize-100-2), 之前跑过一次，这次是复现一下 ☑
    • GPTR-(CCAM-Neg-Last2BLK-1.0T)-(RS-RC-CJ)-(BatchSize-80-2) 06-25 ☑
    • GPTR-(CCAM-Neg-Last2BLK-1.0T)-(RS-RC-CJ)-(BatchSize-128-2) 06-25 ☑
      • 验证大的BatchSize是不是会出现GT_loc达到峰值再下降的原因。
        —
    • GPTR-(CCAM-Neg-Last2BLK-1.0T)-(RS-RC-CJ)-(BatchSize-100-2), 发现有点小差别。。。，Fuck, 再复现一次 ☑
  • (CCAM-Neg-Last2BLK-1.0T) 赋予不同的权重，验证CCAM_loss的权重是不是会出现GT_loc达到峰值再下降的原因。
    • GPTR-(CCAM-Neg-Last2BLK-1.0T-w-0.2)-(RS-RC-CJ)-(BatchSize-64-2) 06-24 ☑
    • GPTR-(CCAM-Neg-Last2BLK-1.0T-w-0.6)-(RS-RC-CJ)-(BatchSize-100-2) 06-24 ☑
    • GPTR-(CCAM-Neg-Last2BLK-1.0T-w-0.6)-(RS-RC-CJ)-(BatchSize-100-2) 06-25 ☑

---

🐯 2022-06-24

🍉 arXiv Paper

• MaskViT: Masked Visual Pre-Training for Video Prediction
• Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation
• NeurIPS_2022_在投_Mutual Information-guided Knowledge Transfer for Novel Class Discovery
• IJCAI2021_Multi-view Feature Augmentation with Adaptive Class Activation Mapping
• ICLR_2020_Self-labelling via simultaneous clustering and representation learning

This site is open source. Improve this page.