加对比学习章节 (#107)

README.md

@@ -39,7 +39,6 @@
 |  ✅  | 2015 |  [ResNet](https://arxiv.org/pdf/1512.03385.pdf) | 构建深层网络都要有的残差连接。         | 80816 ([link](https://www.semanticscholar.org/paper/Deep-Residual-Learning-for-Image-Recognition-He-Zhang/2c03df8b48bf3fa39054345bafabfeff15bfd11d)) |
 |  | 2017 | [MobileNet](https://arxiv.org/pdf/1704.04861.pdf) | 适合终端设备的小CNN             | 8695 ([link](https://www.semanticscholar.org/paper/MobileNets%3A-Efficient-Convolutional-Neural-Networks-Howard-Zhu/3647d6d0f151dc05626449ee09cc7bce55be497e)) |
 | | 2019 | [EfficientNet](https://arxiv.org/pdf/1905.11946.pdf) | 通过架构搜索得到的CNN             | 3426 ([link](https://www.semanticscholar.org/paper/EfficientNet%3A-Rethinking-Model-Scaling-for-Neural-Tan-Le/4f2eda8077dc7a69bb2b4e0a1a086cf054adb3f9)) |
-| ✅ | 2019 | [MoCo](https://arxiv.org/pdf/1911.05722.pdf) | 无监督训练效果也很好             | 2011 ([link](https://www.semanticscholar.org/paper/Momentum-Contrast-for-Unsupervised-Visual-Learning-He-Fan/ec46830a4b275fd01d4de82bffcabe6da086128f)) |
 | | 2021 |  [Non-deep networks](https://arxiv.org/pdf/2110.07641.pdf) | 让不深的网络也能在ImageNet刷到SOTA             | 0 ([link](https://www.semanticscholar.org/paper/Non-deep-Networks-Goyal-Bochkovskiy/0d7f6086772079bc3e243b7b375a9ca1a517ba8b)) |
 
 ### 计算机视觉 - Transformer
@@ -82,6 +81,26 @@
 |        | 2020 | [DETR](https://arxiv.org/pdf/2005.12872.pdf)      | Transformer     | 1053 ([link](https://www.semanticscholar.org/paper/End-to-End-Object-Detection-with-Transformers-Carion-Massa/962dc29fdc3fbdc5930a10aba114050b82fe5a3e)) |
 
 
+### 计算机视觉 - 对比学习
+
+
+| 已录制 | 年份 | 名字                                                         | 简介                 | 引用                                                         |
+| ------ | ---- | ------------------------------------------------------------ | -------------------- | -----------------------------------------------------------: |
+| ✅      | 2018 | [InstDisc](https://arxiv.org/pdf/1805.01978.pdf) | 提出实例判别和memory bank做对比学习            | 885 ([link](https://www.semanticscholar.org/paper/Unsupervised-Feature-Learning-via-Non-parametric-Wu-Xiong/155b7782dbd713982a4133df3aee7adfd0b6b304)) |
+| ✅      | 2018 | [CPC](https://arxiv.org/pdf/1807.03748.pdf) | 对比预测编码，图像语音文本强化学习全都能做             | 2187 ([link](https://www.semanticscholar.org/paper/Representation-Learning-with-Contrastive-Predictive-Oord-Li/b227f3e4c0dc96e5ac5426b85485a70f2175a205)) |
+| ✅      | 2019 | [InvaSpread](https://arxiv.org/pdf/1904.03436.pdf) | 一个编码器的端到端对比学习             | 223 ([link](https://www.semanticscholar.org/paper/Unsupervised-Embedding-Learning-via-Invariant-and-Ye-Zhang/e4bde6fe33b6c2cf9d1647ac0b041f7d1ba29c5b)) |
+| ✅  | 2019 |  [CMC](https://arxiv.org/pdf/1906.05849.pdf) | 多视角下的对比学习         | 780 ([link](https://www.semanticscholar.org/paper/Contrastive-Multiview-Coding-Tian-Krishnan/97f4d09175705be4677d675fa27e55defac44800)) |
+| ✅ | 2019 | [MoCov1](https://arxiv.org/pdf/1911.05722.pdf) | 无监督训练效果也很好             | 2011 ([link](https://www.semanticscholar.org/paper/Momentum-Contrast-for-Unsupervised-Visual-Learning-He-Fan/ec46830a4b275fd01d4de82bffcabe6da086128f)) |
+|  ✅ | 2020 |  [SimCLRv1](https://arxiv.org/pdf/2002.05709.pdf) |  简单的对比学习 (数据增强 + MLP head + 大batch训练久)            | 2958 ([link](https://www.semanticscholar.org/paper/A-Simple-Framework-for-Contrastive-Learning-of-Chen-Kornblith/34733eaf66007516347a40ad5d9bbe1cc9dacb6b)) |
+|  ✅ | 2020 | [MoCov2](https://arxiv.org/pdf/2003.04297.pdf) | MoCov1 + improvements from SimCLRv1             | 725 ([link](https://www.semanticscholar.org/paper/Improved-Baselines-with-Momentum-Contrastive-Chen-Fan/a1b8a8df281bbaec148a897927a49ea47ea31515)) |
+|  ✅ | 2020 |  [SimCLRv2](https://arxiv.org/pdf/2006.10029.pdf) | 大的自监督预训练模型很适合做半监督学习             | 526 ([link](https://www.semanticscholar.org/paper/Big-Self-Supervised-Models-are-Strong-Learners-Chen-Kornblith/3e7f5f4382ac6f9c4fef6197dd21abf74456acd1)) |
+| ✅  | 2020 |  [BYOL](https://arxiv.org/pdf/2006.07733.pdf) | 不需要负样本的对比学习             | 932 ([link](https://www.semanticscholar.org/paper/Bootstrap-Your-Own-Latent%3A-A-New-Approach-to-Grill-Strub/38f93092ece8eee9771e61c1edaf11b1293cae1b)) |
+|  ✅ | 2020 |  [SWaV](https://arxiv.org/pdf/2006.09882.pdf) | 聚类对比学习             | 593 ([link](https://www.semanticscholar.org/paper/Unsupervised-Learning-of-Visual-Features-by-Cluster-Caron-Misra/10161d83d29fc968c4612c9e9e2b61a2fc25842e)) |
+|  ✅ | 2020 |  [SimSiam](https://arxiv.org/pdf/2011.10566.pdf) | 化繁为简的孪生表征学习             | 403 ([link](https://www.semanticscholar.org/paper/Exploring-Simple-Siamese-Representation-Learning-Chen-He/0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d)) |
+| ✅ | 2021 | [MoCov3](https://arxiv.org/pdf/2104.02057.pdf) | 如何更稳定的自监督训练ViT             | 96 ([link](https://www.semanticscholar.org/paper/An-Empirical-Study-of-Training-Self-Supervised-Chen-Xie/739ceacfafb1c4eaa17509351b647c773270b3ae)) |
+|  ✅ | 2021 |  [DINO](https://arxiv.org/pdf/2104.14294.pdf) | transformer加自监督在视觉也很香             | 200 ([link](https://www.semanticscholar.org/paper/Emerging-Properties-in-Self-Supervised-Vision-Caron-Touvron/ad4a0938c48e61b7827869e4ac3baffd0aefab35)) |
+
+
 ### 自然语言处理 - Transformer
 
 | 已录制 | 年份 | 名字                                                         | 简介                 | 引用                                                         |