Facebook researches have developed on a new identification method which uses a subject's pose and other aspects such as hair and clothing to verify identities – sometimes in conjunction with a full frontal face image, sometimes without.Called the Pose Invariant PErson Recognition (PIPER) method, the technique was developed using a new People In Photo Albums (PIPA) dataset, consisting of over 60,000 instances of over 2,000 individuals collected from public Flickr photo albums.Presented as an improvement on the social media giant's Deep Face network, the innovation uses part-level person recognizers or “poselets” to account for pose variations.”While a lot of progress has been made recently in recognition from a frontal face, non-frontal views are a lot more common in photo albums than people might suspect. ߪ Thus the problem of recognizing people from any viewpoint and without the presence of a frontal face or canonical pedestrian pose is important.”Examples of the “poselets” it includes are a hand next to a hip or head-and-shoulders in a back-facing view, or legs of a person walking sideways.Underlining the difficulty of the dataset used, the researchers state that FB's “state-of-the-art” DeepFace system was only able to register only 52% of the instances in it, with an overall accuracy score of just 46.66%.By contrast, PIPER was able to achieve a score of 83.05%. Moreover when a frontal face is available, it improved the accuracy over DeepFace from 89.3% to 93.4%.”We hope our dataset will steer the vision community towards the very important and largely unsolved problem of person recognition in the wild,” wrote the researchers.