Estimating Correspondences of Deformable Objects “In-the-Wild”

Abstract

During the past few years we have witnessed the development of many methodologies for building and fitting Statistical Deformable Models (SDMs). The construction of accurate SDMs requires careful annotation of images with regards to a consistent set of landmarks. However, the manual annotation of a large amount of images is a tedious, laborious and expensive procedure. Furthermore, for several deformable objects, e.g. human body, it is difficult to define a consistent set of landmarks, and, thus, it becomes impossible to train humans in order to accurately annotate a collection of images. Nevertheless, for the majority of objects, it is possible to extract the shape by object segmentation or even by shape drawing. In this paper, we show for the first time, to the best of our knowledge, that it is possible to construct SDMs by putting object shapes in dense correspondence. Such SDMs can be built with much less effort for a large battery of objects. Additionally, we show that, by sampling the dense model, a part-based SDM can be learned with its parts being in correspondence. We employ our framework to develop SDMs of human arms and legs, which can be used for the segmentation of the outline of the human body, as well as to provide better and more consistent annotations for body joints.

The unconstrained ear recognition challenge

“…The group from the Imperial College London (ICL) participated in the UERC with an approach build around Statistical Deformable Models (SDMs) [56] and Inception-ResNets [48]. The SDM was used for dense ear alignment and the Inception-ResNet for descriptor computation.…”

Learning Kinematic Structure Correspondences Using Multi-Order Similarities

“…A key application lies in the robotics field, where the correspondence between a human teacher and a humanoid robot can be established, which allows imitation learning without a pre-defined model of the human. Most conventional correspondence finding methods in the computer vision area are restricted to two static images of the same category [17], [18], [19] or the same object with different poses or views [20], [21], [22], [23]. Local shape feature and graph matching based methods [24], [25], [26] have been researched actively for decades.…”

Improve Accurate Pose Alignment and Action Localization by Dense Pose Estimation

“…Our work in dense body pose estimation networks are inspired by learning with 'Privileged Information' [19], [1], [5], [26], where it is argued that one can simplify training through the use of an 'Intelligent Teacher' that in a way explains the supervision signal, rather than simply penalizing misclassifications. This technique was recently used in deep learning for the task of image classification [5].…”