DGPose: Deep Generative Models for Human Body Analysis

Deep generative modelling for human body analysis is an emerging problem with many interesting applications. However, the latent space learned by such approaches is typically not interpretable, resulting in less flexibility. In this work, we present deep generative models for human body analysis in...

Fuld beskrivelse

Bibliografiske detaljer
Main Authors: de Bem, R, Ghosh, A, Ajanthan, T, Miksik, O, Boukhayma, A, Siddharth, N, Torr, P
Format: Journal article
Udgivet: Springer 2020
_version_ 1826286541172899840
author de Bem, R
Ghosh, A
Ajanthan, T
Miksik, O
Boukhayma, A
Siddharth, N
Torr, P
author_facet de Bem, R
Ghosh, A
Ajanthan, T
Miksik, O
Boukhayma, A
Siddharth, N
Torr, P
author_sort de Bem, R
collection OXFORD
description Deep generative modelling for human body analysis is an emerging problem with many interesting applications. However, the latent space learned by such approaches is typically not interpretable, resulting in less flexibility. In this work, we present deep generative models for human body analysis in which the body pose and the visual appearance are disentangled. Such a disentanglement allows independent manipulation of pose and appearance, and hence enables applications such as posetransfer without specific training for such a task. Our proposed models, the Conditional-DGPose and the Semi-DGPose, have different characteristics. In the first, body pose labels are taken as conditioners, from a fully-supervised training set. In the second, our structured semi-supervised approach allows for pose estimation to be performed by the model itself and relaxes the need for labelled data. Therefore, the Semi-DGPose aims for the joint understanding and generation of people in images. It is not only capable of mapping images to interpretable latent representations but also able to map these representations back to the image space. We compare our models with relevant baselines, the ClothNet-Body and the Pose Guided Person Generation networks, demonstrating their merits on the Human3.6M, ChictopiaPlus and DeepFashion benchmarks.
first_indexed 2024-03-07T01:45:18Z
format Journal article
id oxford-uuid:9834505c-1cfb-495f-a46d-2ddc04d64861
institution University of Oxford
last_indexed 2024-03-07T01:45:18Z
publishDate 2020
publisher Springer
record_format dspace
spelling oxford-uuid:9834505c-1cfb-495f-a46d-2ddc04d648612022-03-27T00:05:31ZDGPose: Deep Generative Models for Human Body AnalysisJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:9834505c-1cfb-495f-a46d-2ddc04d64861Symplectic ElementsSpringer2020de Bem, RGhosh, AAjanthan, TMiksik, OBoukhayma, ASiddharth, NTorr, PDeep generative modelling for human body analysis is an emerging problem with many interesting applications. However, the latent space learned by such approaches is typically not interpretable, resulting in less flexibility. In this work, we present deep generative models for human body analysis in which the body pose and the visual appearance are disentangled. Such a disentanglement allows independent manipulation of pose and appearance, and hence enables applications such as posetransfer without specific training for such a task. Our proposed models, the Conditional-DGPose and the Semi-DGPose, have different characteristics. In the first, body pose labels are taken as conditioners, from a fully-supervised training set. In the second, our structured semi-supervised approach allows for pose estimation to be performed by the model itself and relaxes the need for labelled data. Therefore, the Semi-DGPose aims for the joint understanding and generation of people in images. It is not only capable of mapping images to interpretable latent representations but also able to map these representations back to the image space. We compare our models with relevant baselines, the ClothNet-Body and the Pose Guided Person Generation networks, demonstrating their merits on the Human3.6M, ChictopiaPlus and DeepFashion benchmarks.
spellingShingle de Bem, R
Ghosh, A
Ajanthan, T
Miksik, O
Boukhayma, A
Siddharth, N
Torr, P
DGPose: Deep Generative Models for Human Body Analysis
title DGPose: Deep Generative Models for Human Body Analysis
title_full DGPose: Deep Generative Models for Human Body Analysis
title_fullStr DGPose: Deep Generative Models for Human Body Analysis
title_full_unstemmed DGPose: Deep Generative Models for Human Body Analysis
title_short DGPose: Deep Generative Models for Human Body Analysis
title_sort dgpose deep generative models for human body analysis
work_keys_str_mv AT debemr dgposedeepgenerativemodelsforhumanbodyanalysis
AT ghosha dgposedeepgenerativemodelsforhumanbodyanalysis
AT ajanthant dgposedeepgenerativemodelsforhumanbodyanalysis
AT miksiko dgposedeepgenerativemodelsforhumanbodyanalysis
AT boukhaymaa dgposedeepgenerativemodelsforhumanbodyanalysis
AT siddharthn dgposedeepgenerativemodelsforhumanbodyanalysis
AT torrp dgposedeepgenerativemodelsforhumanbodyanalysis