Long-tailed recognition學習目錄

為何稱為Long-tailed recognition？在日常生活中，很多的資料分布都遵循長尾分布。同樣地，在電腦視覺中的dataset也會出現這樣的分布情形。例如某些類別的圖片很多，某些類別的圖片特別少。這樣的情形會造成在圖片較少的類別上，辨識準確度較差的情形產生。Long-tailed recognition的研究旨在減緩dataset不平衡所造成的問題。

20210621更新：Micro avg和Marco avg

F1 Score
Micro avg F1: 較考慮模型「判斷力」較高的class
Marco avg F1: 考慮imblanced的class

多分类条件下分类指标：Macro-average和Micro-average，以及Weighted-average_jmhIcoding-CSDN博客

Macro- or micro-average for imbalanced class problems

#轉自https://datascience.stackexchange.com/a/72684/115316
F1 'micro' - the micro weighs each sample equally
class 1 accounted for 40% of the data, F1 for this class is 0.8      class 2 accounted for 60% of the data, F1 for this class is 0.2
0.8 x 40% + 0.2 x 60% = 0.44
F1 'macro' - the macro weighs each class equally
class 1: the F1 result = 0.8 for class 1 F1 result = 0.2 for class 2.
We do the usual arthmetic average:      (0.8 + 0.2) / 2 = 0.5
It would be the same no matter how the samples are split between two classes. The choice depends on what you want to achieve. If you're worried about class imbalances, I'd suggest using a 'macro'.

20210218更新：A Simple and Effective Use of Object-Centric Images for Long-Tailed Object Detection

主要想法是使用Object-Centric的圖片進行資料增強，進而提升偵測結果，實驗結果看起來是目前常見方法中最佳的，優於Balanced Meta-softmax

RSG: A Simple but Effective Module for Learning Imbalanced Datasets (CVPR 2021)

Jianf-Wang/RSG

RSG: A Simple but Effective Module for Learning Imbalanced Datasets

A Simple and Effective Use of Object-Centric Images for Long-Tailed Object Detection

Balanced Meta-Softmax for Long-Tailed Visual Recognition

jiawei-ren/BalancedMetaSoftmax-Classification

Balanced Meta-Softmax for Long-Tailed Visual Recognition

Equalization Loss for Long-Tailed Object Recognition

Overcoming Classifier Imbalance for Long-tail Object Detection with Balanced Group Softmax

Seesaw Loss for Long-Tailed Instance Segmentation

Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect

Forest R-CNN: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentation