human pose estimation
2D Pose estimation主要面臨的困難:遮擋、復雜背景、光照、真實世界的復雜姿態、人的尺度不一、拍攝角度不固定等。
單人姿態估計
傳統方法:基於Pictorial Structures, DPM
? 基於深度學習的算法包括直接回歸坐標(Deep Pose)和通過熱力圖回歸坐標(CPM, Hourlgass)
目前單人姿態估計,主流算法是基於Hourlgass各種更改結構的算法。
多人姿態估計
二維圖像姿態估計基於CNN的多人姿態估計方法,通常有2個思路(Bottom-Up Approaches和Top-Down Approaches):
(1)Top-Down Approaches,即two-step framework,就是先進行行人檢測,得到邊界框,然後在每一個邊界框中檢測人體關鍵點,連接成一個人形,缺點就是受檢測框的影響太大,漏檢,誤檢,IOU大小等都會對結果有影響,算法包括RMPE、Mask-RCNN 等。
(2)Bottom-Up Approaches,即part-based framework,就是先對整個圖片進行每個人體關鍵點部件的檢測,再將檢測到的部件拼接成一個人形,缺點就是會將不同人的不同部位按一個人進行拼接,代表方法就是openpose、DeepCut 、PAFs。
tricks
- 采用多尺度,多分辨率的網絡結構
- 采用基於Residual Block來構建網絡
- 擴大感受野(large kernel, dilation convolution, Spatial Transformer Network、hourglass module)
- 預處理很重要(將人放在輸入圖片的中心,人的尺度盡量歸一化到統一尺度,對圖片進行翻轉、旋轉)
- 後處理同樣重要
openpose源碼中subset輸出的關鍵點順序是:1鼻子,2脖子,3右肩,4右肘,5右腕,6左肩,7左肘,8左腕,9右髖,10右膝,11右踝,12左髖,13左膝,14左踝,15左眼,16右眼,17左耳,18右耳,19 pt19
輸出格式;https://www.aiuai.cn/aifarm712.html
CPM
paper:
https://blog.csdn.net/shenxiaolu1984/article/details/51094959
openPose
GitHub:
Realtime_Multi-Person_Pose_Estimation
https://github.com/CMU-Perceptual-Computing-Lab/openpose
配置:
https://blog.csdn.net/lgh0824/article/details/75949477
https://blog.csdn.net/hk121/article/details/83537350
openPose解析
https://blog.csdn.net/qq_27158179/article/details/82717821
https://www.jianshu.com/c/8602d176d8ea?utm_source=desktop&utm_medium=notes-included-collection
https://zhuanlan.zhihu.com/p/48507352
[OpenPose翻譯] Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields ?原文翻譯(註釋版)
https://blog.csdn.net/kenllf/article/details/79702078
Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields ? 實時多人人體姿態估計論文原理講解
https://blog.csdn.net/Lin_xiaoyi/article/details/78838393
https://blog.csdn.net/yxr403614258/article/details/77977330
Paper reading: Realtime Multi-person 2D Pose estimation using Part Affinity Fields(1)
https://blog.csdn.net/yengjie2200/article/details/68064095
openpose實驗總結
https://blog.csdn.net/qq_20657717/article/details/81707746
膚色檢測
https://blog.csdn.net/yangtrees/article/details/8269984
基於顏色檢測物體
http://www.cnblogs.com/wangxinyu0628/p/5928824.html
項目編譯:
https://blog.csdn.net/zb1165048017/article/details/82115724
姿態估計的應用:
https://blog.csdn.net/itchosen/article/details/77200384
https://blog.csdn.net/shenxiaolu1984/article/details/51094959
https://blog.csdn.net/yeahDeDiQiZhang/article/details/78131566
https://www.cnblogs.com/JillBlogs/p/9098989.html
Stacked Hourglass算法詳解
https://blog.csdn.net/shenxiaolu1984/article/details/51428392
代碼閱讀】OpenPose(Pytorch Realtime Multi-Person Pose Estimation)
https://blog.csdn.net/a529975125/article/details/80991781
pytorch千千問
https://blog.csdn.net/daniaokuye/article/details/78851479
human pose estimation