@AikenHong2021 OWL

分析现有的OWL特点，和当前自己的研究做一个区分，也汲取一下别人的研究的要点，

Reference

1. arxiv @ self-supervised feature improve open-world learning
2. arxiv @ open-world semi-supervised learning
3. arxiv @ open-world learning without labels
4. arxiv @ unseen class discovery in open-world classification
5. arxiv @ Open-World Active Learning with Stacking Ensemble for Self-Driving Cars
6. www @ open-world learning and application to product classification
7. cvpr @ open world composition zero-shot learning
8. cvpr @ Towards Open World Object Detection
9. cvpr](https://openaccess.thecvf.com/content_CVPR_2019/papers/Liu_Large-Scale_Long-Tailed_Recognition_in_an_Open_World_CVPR_2019_paper.pdf)) @ Large-Scale Long-Tailed Recognition in an Open World

Conclusion

Papers

Mulit Open world Learning Definition

拒绝未见过的类的实例，逐步学习新的类扩展现有模型

:zap: Large-Scale Long-Tailed Recognition in an Open World

Large-Scale Long-Tailed Recognition in an Open World (liuziwei7.github.io)

@Aiken 2021

框架撞车系列，主要看看这一篇论文中怎么解决如下的问题👇，并从中借鉴和优化的我框架设计

思路分析

Motivation

模型实现的主要的两个TASK：

1. Open Set Learning ： 在没有明确监督的时候，将尚未引入的目标类别识别为未知
2. Incremental Learning：类别增量学习

实现这两个问题的主要思路：

• 自动标注：借鉴RPN的class-agnostic，以及检测和分类的显著性指标的差异，找到并自动标注NewClass
• 对比聚类：使用prototype feature来进行聚类，同时计算Distance损失
  it seems like contain a unknown prototype.
• energy based：亥姆霍兹自由能公式？

ENERGY BASED