The visual analysis of the body, including posture and movement, represents a rich source of information for numerous biomedical applications. For example: a simple image of the body reveals much about how athletic or sedentary a person’s lifestyle is; a correct walking pattern can indicate recovery after a rehabilitation period; determining whether a person is sitting correctly on a chair can help prevent back injuries; and keeping the right distance and body orientation with respect to other people can signal compliance with social norms. All of these analyses are currently carried out by medical specialists or social psychologists, which requires a large amount of time and results in diagnoses which are highly dependent on the expert who evaluates the images. At the same time, computer vision has advanced greatly in the analysis of the human body, due to the development of deep learning technologies able to provide robust descriptions of the human body in quantitative terms.
This Research Topic aims to highlight the latest research on cutting-edge technologies for the automated analysis of the human body which are visual (i.e., without involving body imaging, so excluding techniques based on X-ray and MRI imaging, etc.) and which do not require any sensors beyond simple RGB and/or depth cameras, possibly paired with wearable sensors. In particular, the focus of submissions should be on the precision and robustness of the measurements obtained, and on their confidence. These aspects are of primary importance in the context of biomedical applications, in which such measurements must be noise-free or within specific ranges of variability.
This Research Topic is intended to present a focus on techniques that take the entire body as their input (rather than, for example, using hand gesture recognition or facial expression analysis). Topics of interest include, but are not limited to:
- (markerless) body pose estimation
- 3D, multicamera, body pose estimation
- body pose forecasting and anomaly detection
- gait analysis
- social signals for the (whole) body
- proxemics and body kinesics
- body anthropometrics from body images and video
- applications in rehabilitation and psychiatry
- issues of ethics and privacy
- explainable/transparent machine learning applications
Topic Editor Prof Marco Cristani is a co-founder of the private company Humatics S.r.l., but declares no competing interests with regards to the Research Topic subject.
The visual analysis of the body, including posture and movement, represents a rich source of information for numerous biomedical applications. For example: a simple image of the body reveals much about how athletic or sedentary a person’s lifestyle is; a correct walking pattern can indicate recovery after a rehabilitation period; determining whether a person is sitting correctly on a chair can help prevent back injuries; and keeping the right distance and body orientation with respect to other people can signal compliance with social norms. All of these analyses are currently carried out by medical specialists or social psychologists, which requires a large amount of time and results in diagnoses which are highly dependent on the expert who evaluates the images. At the same time, computer vision has advanced greatly in the analysis of the human body, due to the development of deep learning technologies able to provide robust descriptions of the human body in quantitative terms.
This Research Topic aims to highlight the latest research on cutting-edge technologies for the automated analysis of the human body which are visual (i.e., without involving body imaging, so excluding techniques based on X-ray and MRI imaging, etc.) and which do not require any sensors beyond simple RGB and/or depth cameras, possibly paired with wearable sensors. In particular, the focus of submissions should be on the precision and robustness of the measurements obtained, and on their confidence. These aspects are of primary importance in the context of biomedical applications, in which such measurements must be noise-free or within specific ranges of variability.
This Research Topic is intended to present a focus on techniques that take the entire body as their input (rather than, for example, using hand gesture recognition or facial expression analysis). Topics of interest include, but are not limited to:
- (markerless) body pose estimation
- 3D, multicamera, body pose estimation
- body pose forecasting and anomaly detection
- gait analysis
- social signals for the (whole) body
- proxemics and body kinesics
- body anthropometrics from body images and video
- applications in rehabilitation and psychiatry
- issues of ethics and privacy
- explainable/transparent machine learning applications
Topic Editor Prof Marco Cristani is a co-founder of the private company Humatics S.r.l., but declares no competing interests with regards to the Research Topic subject.