Uploaded January 2020 | Updated September 2026, 2 days ago
Second Prize in our 2019 ZEISS Photography Competition. eng.cam.ac.uk/news/winners-department-s-photography-competition-unveiled
Human body parsing is the process of segmenting an image or video of a person into semantically meaningful parts. The system Dr James Charles and his team designed is capable of automatically parsing multiple people in video and decomposing them into clothing items and body parts.
Here they show the output from their real-time deep learning based system, when applied to an image of a dancer. Different colours represent different parts; on the left are clothing items and on the right body parts are shown.
When applied to video, tracking of individual people and their parts is accomplished by matching segments together along a temporal sequence. Their algorithm is quite efficient and able to run in real-time live on a mobile phone. Such systems have many applications e.g. making self-driving cars aware of the motion and behaviour of pedestrians.
Second Prize in our 2019 ZEISS Photography Competition. eng.cam.ac.uk/news/winners-department-s-photography-competition-unveiled
Human body parsing is the process of segmenting an image or video of a person into semantically meaningful parts. The system Dr James Charles and his team designed is capable of automatically parsing multiple people in video and decomposing them into clothing items and body parts.
Here they show the output from their real-time deep learning based system, when applied to an image of a dancer. Different colours represent different parts; on the left are clothing items and on the right body parts are shown.
When applied to video, tracking of individual people and their parts is accomplished by matching segments together along a temporal sequence. Their algorithm is quite efficient and able to run in real-time live on a mobile phone. Such systems have many applications e.g. making self-driving cars aware of the motion and behaviour of pedestrians.
