Estimation of partially occluded 2D human joints with a Bayesian approach

2021-07-01
Dursun, Ahmet Anıl
Tuncer, Temel Engin
Show full item record
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
151
views
0
downloads
Human pose estimation is an important problem which has found applications in different fields. One problem in this context is to estimate the human joints or keypoints under occlusion. Most of the methods in the literature use CNN structures to estimate keypoint positions. In this paper, a Bayesian Approach for occluded Keypoint Estimation, BAKE, is presented to complete the missing human pose elements. A state-of-the-art CNN-based pose estimation algorithm Openpose [1] is used for detecting the visible joints. Then a Bayesian pose estimation is applied for the missing 2D pose elements. A Gaussian model composed of the length and angle parameters of a 13-keypoint skeleton model to encode the 2D human body pose is presented. Model parameters are obtained from the global and local statistics. Global statistics are extracted from COCO database [2] whereas the local statistics are obtained from the non-occluded initial video frames under consideration. The local and global statistics are combined and updated by using the action-specific body relations. A new predictability score is proposed to determine the confidence of completing the missing human joints. This confidence score is used to develop a hybrid technique which combines the predictions of the Openpose and the proposed method, BAKE. The proposed approach can be seen as an alternative for a 3D CNN structure which is trying to model the time-dependent sequence of events. Several experiments are performed to compare the outputs of the Openpose, BAKE, and the hybrid approach. It is shown that BAKE outperforms the Openpose estimates in general and the hybrid method generates a slight improvement over the BAKE algorithm.
https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85104381276&origin=inward
https://hdl.handle.net/11511/90826
Digital Signal Processing: A Review Journal
Department of Electrical and Electronics Engineering, Article

Suggestions

Development of synthetic and real-world pose estimation dataset to be used in human tracking system
Ersoy, Mustafa; Koku, Ahmet Buğra; Department of Mechanical Engineering (2022-4-29)
In this study, we propose an extendable, synthetic human pose estimation dataset named “Metupose”. Pose estimation aims to determine the pose of a person by detecting joints in an image or video. Dataset was created in Blender 3D software and with varying human objects and environment. It is also used to enhance the accuracy of pose estimation models in the literature. Metupose dataset contains 178000 images. Images have 1 to 4 people in it, where there is a total of 402000 people exist in these images. Whe...
Bayesian model pooling for the analysis of generalized linear models with missing-not-at-random covariates
Çiftçi, Sezgin; Kalaylıoğlu Akyıldız, Zeynep Işıl; Department of Statistics (2020-9)
Missing data in modeling is one of the main problems of researchers. If the missingness mechanism is related to the value of the variable itself, then the missing is defined as not-at-random (MNAR). Additionally, finding the best model in statistics has always been the focus of researchers. However, it's known from examples that there may be any other model with the same best fit but has different estimated coefficients. Hence, an uncertainty about model selection that may be problematic in estimation espec...
Interpretation of searches for supersymmetry with simplified models
Chatrchyan, S.; et. al. (American Physical Society (APS), 2013-09-01)
The results of searches for supersymmetry by the CMS experiment are interpreted in the framework of simplified models. The results are based on data corresponding to an integrated luminosity of 4.73 to 4.98 fb(-1). The data were collected at the LHC in proton-proton collisions at a center-of-mass energy of 7 TeV. This paper describes the method of interpretation and provides upper limits on the product of the production cross section and branching fraction as a function of new particle masses for a number o...
Optimum design of steel structures via differential evolution algorithm and application programming interface of SAP2000
Dedekargınoğlu, Özgür; Hasançebi, Oğuzhan; Department of Civil Engineering (2012)
The objective of this study is to investigate the use and efficiency of Differential Evolution (DE) method on structural optimization. The solution algorithm developed with DE is computerized into software called SOP2011 using VB.NET. SOP2011 is automated to achieve size optimum design of steel structures consisting of 1-D elements such as trusses and frames subjected to design provisions according to ASD-AISC (2010) and LRFD-AISC (2010). SOP2011 works simultaneously with the structural analysis and design ...
A powerful approach for effective finding of significantly differentially expressed genes
Abul, Osman; Alhajj, Reda; Polat, Faruk (Institute of Electrical and Electronics Engineers (IEEE), 2006-07-01)
The problem of identifying significantly differentially expressed genes for replicated microarray experiments is accepted as significant and has been tackled by several researchers. Patterns from Gene Expression (PaGE) and q-values are two of the well-known approaches developed to handle this problem. This paper proposes a powerful approach to handle this problem. We first propose a method for estimating the prior probabilities used in the first version of the PaGE algorithm. This way, the problem definitio...
Citation Formats
A. A. Dursun and T. E. Tuncer, “Estimation of partially occluded 2D human joints with a Bayesian approach,” Digital Signal Processing: A Review Journal, pp. 0–0, 2021, Accessed: 00, 2021. [Online]. Available: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85104381276&origin=inward.
METU IIS - Integrated Information Service open@metu.edu.tr