Improving Content-Invariance in Gated Autoencoders for 2D and 3D Object Rotation

نویسندگان

  • Stefan Lattner
  • Maarten Grachten
چکیده

Content-invariance in mapping codes learned by GAEs is a useful feature for various relation learning tasks. In this paper we show that the content-invariance of mapping codes for images of 2D and 3D rotated objects can be substantially improved by extending the standard GAE loss (symmetric reconstruction error) with a regularization term that penalizes the symmetric cross-reconstruction error. This error term involves reconstruction of pairs with mapping codes obtained from other pairs exhibiting similar transformations. Although this would principally require knowledge of the transformations exhibited by training pairs, our experiments show that a bootstrapping approach can sidestep this issue, and that the regularization term can effectively be used in an unsupervised setting.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Performance evaluation of gated volumetric modulated arc therapy

Background: Aim of this study is to evaluate the accuracy of the gated volumetric modulated arc therapy (VMAT/RapidArc) using 2D planar dosimetry, DynaLog files and COMPASS 3D dosimetry system. Materials and Methods: Pre-treatment quality assurance of 10 gated VMAT plans was verified using 2D array and COMPASS 3D dosimetry system. Advantage of COMPASS over 2D planar is that it provides the clin...

متن کامل

3D Object Registration and Recognition using Range Images

In recent years, retrieving semantic information from digital cameras, for instance object recognition, has become one of the hottest topics of computer vision. Since the boundaries of this problem range from recognizing objects in a range image to estimating the pose of an object from an image sequence, a variety of studies exist in the literature. Before remembering the previous approaches on...

متن کامل

A novel Local feature descriptor using the Mercator projection for 3D object recognition

Point cloud processing is a rapidly growing research area of computer vision. Introducing of cheap range sensors has made a great interest in the point cloud processing and 3D object recognition. 3D object recognition methods can be divided into two categories: global and local feature-based methods. Global features describe the entire model shape whereas local features encode the neighborhood ...

متن کامل

Why The Brain Separates Face Recognition From Object Recognition

Many studies have uncovered evidence that visual cortex contains specialized regions involved in processing faces but not other object classes. Recent electrophysiology studies of cells in several of these specialized regions revealed that at least some of these regions are organized in a hierarchical manner with viewpointspecific cells projecting to downstream viewpoint-invariant identity-spec...

متن کامل

O-16: Comparison of Pre-Antral Follicle Culture Development during 2 Dimensional and 3 Dimensional Culture Systems

Background: Setting up an in vitro follicle culture system that resembles in vivo ovary condition has high value in research. Additionally, expression evaluation of folliculogenesis involved genes could lead us to the designing of better culture system. Materials and Methods: ovaries of 12-day-old female NMRI mice were removed, 100-130 μm pre-antral follicles were mechanically isolated from fre...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • CoRR

دوره abs/1707.01357  شماره 

صفحات  -

تاریخ انتشار 2017