Crafting a multi-task CNN for viewpoint estimation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année :

Crafting a multi-task CNN for viewpoint estimation

Francisco Massa
Renaud Marlet
Mathieu Aubry
  • Fonction : Auteur
  • PersonId : 964162


Convolutional Neural Networks (CNNs) were recently shown to provide state-of-the- art results for object category viewpoint estimation. However different ways of formulat- ing this problem have been proposed and the competing approaches have been explored with very different design choices. This paper presents a comparison of these approaches in a unified setting as well as a detailed analysis of the key factors that impact perfor- mance. Followingly, we present a new joint training method with the detection task and demonstrate its benefit. We also highlight the superiority of classification approaches over regression approaches, quantify the benefits of deeper architectures and extended training data, and demonstrate that synthetic data is beneficial even when using ImageNet training data. By combining all these elements, we demonstrate an improvement of ap- proximately 5% mAVP over previous state-of-the-art results on the Pascal3D+ dataset [28]. In particular for their most challenging 24 view classification task we improve the results from 31.1% to 36.1% mAVP.
Fichier principal
Vignette du fichier
BMVC-2016-Massa-et-al.pdf (206.21 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-01743267 , version 1 (26-03-2018)



Francisco Massa, Renaud Marlet, Mathieu Aubry. Crafting a multi-task CNN for viewpoint estimation. British Machine Vision Conference (BMVC 2016), Sep 2016, York, United Kingdom. ⟨10.5244/C.30.91⟩. ⟨hal-01743267⟩
96 Consultations
527 Téléchargements



Gmail Facebook Twitter LinkedIn More