How to learn from unlabeled volume data: Self-Supervised 3D Context Feature Learning
Author：Maximilian Blendowski（University of L¨ubeck, Germany）
提出的2.5D的自监督方法，预测不同slice之间某两点的相对位移。使用两个预测网络：直接数值预测和热力图方法。主要用于胸片数据。
Scaling and Benchmarking Self-Supervised Visual Representation Learning
Author：Priya Goyal 等（Facebook）
提出了几个self-supervised 的trick：
-
increasing the size of pre-training data improves the transfer learning performance for both the Jigsaw and Colorization methods
-
Pre-train 的模型越大越好（resnet50>Alexnet）
Self-supervised Spatiotemporal Feature Learning by Video Geometric Transformations
- 处理数据：video
- 方法：
- a set of pre-designed geometric transformations (e.g. rotating 0°, 90°, 180°, and 270°) are applied to each video
- 预测 transformations (e.g. rotations)
Mix-and-Match Tuning for Self-Supervised Semantic Segmentation（2017CVPR）
author: Xiaohang Zhan(港中文)
![]()
- 分为3steps：1）pre-train learning by colorization；2)M&M tuning； 3）target segmentation task
- M&M tuning：1）对图像采取图像块，去除严重重叠的图像块，根据标记的图像真值提取图像块对应的 unique class labels（比如车、人） ，将这些图像块全部混合在一起。2) fine-tuning the network by triplet loss
Boosting Self-Supervised Learning via Knowledge Transfer
- 整体框架图
- （a）中的pretext task是Jigsaw++ task
![]()
- In Jigsaw++, we replace a random number of tiles in the grid (up to 2) with (occluding) tiles from another random image
DeepCluster
使用聚类对每个类产生伪标签，用伪标签作为预训练。