خانک ای ار

بورسیه دکتری برق/الکترونیک/مخابرات|کامپیوتر/فناوری اطلاعات|ریاضی/آمار در فرانسه

بورسیه دکتری برق/الکترونیک/مخابرات|کامپیوتر/فناوری اطلاعات|ریاضی/آمار در فرانسه

The field of computer vision has recently experienced a drastic transformation since the adoption of machine learning techniques, in particular deep learning, to solve problems such as recognition and detection in images and video. However, although the effects and benefits of applying these techniques to 3D and 4D (dynamic scenes) modelling have been anticipated, they have yet to be fully and formally investigated. In particular it is expected that they increase model precision with learned priors, simplify the acquisition process by exploiting learned information and reduce data sizes with learned statistical models. However, successful learning techniques in 2D computer vision, e.g. convolutional deep networks, do not easily generalise to 3D and 4D data since the regular grid assumption with 2D images does not have a straightforward equivalent in 3D-4D. In order to benefit from these techniques, new representations that can learn moving shape properties must be proposed. This area is very promising and expected to have a significant impact on 3D and 4D shape analysis over the coming years.

Recent related works in the fields of machine learning, computer vision and computer graphics have explored two main strategies with 3D geometry. A first spatial or extrinsic strategy consists in embedding the shape geometry into Euclidean structures over which standard CNN tools can be applied. This can be 2D structures, as with depth image projections [1], or 3D structure, for instance voxels [2]. A second category considers instead spectral or intrinsic representations that provide Fourier like decompositions of shapes over spectral domains [3,4]. These eigen decompositions enable then well defined convolutions and multi-scale analysis over non Euclidean manifolds, but the spectral representation does not generalise well across different shapes. A third approach defines local activation functions over shapes by relying on filters with learned weights and shapes over locally defined coordinate systems [5].

[۱] Dense Human Body Correspondences Using Convolutional Networks,

  1. Wei, Q. Huang, D. Ceylan, E. Vouga, H. Li, CVPR 2016.
[۲] ۳D ShapeNets: A Deep Representation for Volumetric Shape Modeling,

  1. Wu, S. Song, A. Khosla, F. Yu, L. Zhang, X. Tang and J. Xiao, CVPR 2015.
[۳] Learning shape correspondence with anisotropic convolutional neural networks,

  1. Boscaini, J .Masci, E. Rodolà, M. Bronstein, NIPS, 2016.
[۴] Spectral Networks and Deep Locally Connected Networks on Graphs,

  1. Bruna, W. Zaremba, A. Szlam, Y. Lecun, ICLR 2014.
[۵] Geometric deep learning on graphs and manifolds using mixture model CNNs

Monti, F.; Boscaini, D.; Masci, J.; Rodolà, E.; Svoboda, J. & Bronstein, M.

ArXiv pre-print, November 2015.

کانال پل کنکوری ارشد دکترا
مطالب مرتبط:

پاسخ دهید

نشانی ایمیل شما منتشر نخواهد شد. بخش‌های موردنیاز علامت‌گذاری شده‌اند *