PA-FlowNet: Pose-Auxiliary Optical Flow Network for Spacecraft Relative Pose Estimation

Chen, Zhiyu; Chen, Po-Heng; Chen, Kuan-Wen; Chan, Chen-Yu

doi:10.1109/icpr48806.2021.9413201

Cited by 2 publications

(1 citation statement)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AlphaPose achieved a performance of 82.1 mAP on MPII, while 75 mAP on COCO dataset. AlphaPose provides an additional pose tracker [25] to track the same person across multiple sequential frames. For anthropometric soft biometrics estimation, AlphaPose can provide 17 body landmarks along with their confidence scores.…”

Section: Alphaposementioning

confidence: 99%

Estimating Anthropometric Soft Biometrics: An Empirical Method

Hassan,

Husnain Raza Sherazi,

Ali

et al. 2023

Intelligent Automation &Amp; Soft Computing

View full text Add to dashboard Cite

Following the success of soft biometrics over traditional biometrics, anthropometric soft biometrics are emerging as candidate features for recognition or retrieval using an image/video. Anthropometric soft biometrics uses a quantitative mode of annotation which is a relatively better method for annotation than qualitative annotations adopted by traditional biometrics. However, one of the most challenging tasks is to achieve a higher level of accuracy while estimating anthropometric soft biometrics using an image or video. The level of accuracy is usually affected by several contextual factors such as overlapping body components, an angle from the camera, and ambient conditions. Exploring and developing such a collection of anthropometric soft biometrics that are less sensitive to contextual factors and are relatively easy to estimate using an image or video is a potential research domain and it has a lot of value for improved recognition or retrieval. For this purpose, anthropometric soft biometrics, which are originally geometric measurements of the human body, can be computed with ease and higher accuracy using landmarks information from the human body. To this end, several key contributions are made in this paper; i) summarizing a range of human body pose estimation tools used to localize dozens of different multi-modality landmarks from the human body, ii) a critical evaluation of the usefulness of anthropometric soft biometrics in recognition or retrieval tasks using state of the art in the field, iii) an investigation on several benchmark human body anthropometric datasets and their usefulness for the evaluation of any anthropometric soft biometric system, and iv) finally, a novel bag of anthropometric soft biometrics containing a list of anthropometrics is presented those are practically possible to measure from an image or video. To the best of our knowledge, anthropometric soft biometrics are potential features for improved seamless recognition or retrieval in both constrained and unconstrained scenarios and they also minimize the approximation level of feature value estimation than traditional biometrics. In our opinion, anthropometric soft biometrics constitutes a practical approach for recognition using closed-circuit television (CCTV) or retrieval from the image dataset, while the bag of anthropometric soft biometrics presented contains a potential collection of biometric features which are less sensitive to contextual factors.

show abstract

Section: Alphaposementioning

confidence: 99%