Header logo is ps


2019


Towards Geometric Understanding of Motion
Towards Geometric Understanding of Motion

Ranjan, A.

University of Tübingen, December 2019 (phdthesis)

Abstract

The motion of the world is inherently dependent on the spatial structure of the world and its geometry. Therefore, classical optical flow methods try to model this geometry to solve for the motion. However, recent deep learning methods take a completely different approach. They try to predict optical flow by learning from labelled data. Although deep networks have shown state-of-the-art performance on classification problems in computer vision, they have not been as effective in solving optical flow. The key reason is that deep learning methods do not explicitly model the structure of the world in a neural network, and instead expect the network to learn about the structure from data. We hypothesize that it is difficult for a network to learn about motion without any constraint on the structure of the world. Therefore, we explore several approaches to explicitly model the geometry of the world and its spatial structure in deep neural networks.

The spatial structure in images can be captured by representing it at multiple scales. To represent multiple scales of images in deep neural nets, we introduce a Spatial Pyramid Network (SpyNet). Such a network can leverage global information for estimating large motions and local information for estimating small motions. We show that SpyNet significantly improves over previous optical flow networks while also being the smallest and fastest neural network for motion estimation. SPyNet achieves a 97% reduction in model parameters over previous methods and is more accurate.

The spatial structure of the world extends to people and their motion. Humans have a very well-defined structure, and this information is useful in estimating optical flow for humans. To leverage this information, we create a synthetic dataset for human optical flow using a statistical human body model and motion capture sequences. We use this dataset to train deep networks and see significant improvement in the ability of the networks to estimate human optical flow.

The structure and geometry of the world affects the motion. Therefore, learning about the structure of the scene together with the motion can benefit both problems. To facilitate this, we introduce Competitive Collaboration, where several neural networks are constrained by geometry and can jointly learn about structure and motion in the scene without any labels. To this end, we show that jointly learning single view depth prediction, camera motion, optical flow and motion segmentation using Competitive Collaboration achieves state-of-the-art results among unsupervised approaches.

Our findings provide support for our hypothesis that explicit constraints on structure and geometry of the world lead to better methods for motion estimation.

PhD Thesis [BibTex]

2019

PhD Thesis [BibTex]

2016


Non-parametric Models for Structured Data and Applications to Human Bodies and Natural Scenes
Non-parametric Models for Structured Data and Applications to Human Bodies and Natural Scenes

Lehrmann, A.

ETH Zurich, July 2016 (phdthesis)

Abstract
The purpose of this thesis is the study of non-parametric models for structured data and their fields of application in computer vision. We aim at the development of context-sensitive architectures which are both expressive and efficient. Our focus is on directed graphical models, in particular Bayesian networks, where we combine the flexibility of non-parametric local distributions with the efficiency of a global topology with bounded treewidth. A bound on the treewidth is obtained by either constraining the maximum indegree of the underlying graph structure or by introducing determinism. The non-parametric distributions in the nodes of the graph are given by decision trees or kernel density estimators. The information flow implied by specific network topologies, especially the resultant (conditional) independencies, allows for a natural integration and control of contextual information. We distinguish between three different types of context: static, dynamic, and semantic. In four different approaches we propose models which exhibit varying combinations of these contextual properties and allow modeling of structured data in space, time, and hierarchies derived thereof. The generative character of the presented models enables a direct synthesis of plausible hypotheses. Extensive experiments validate the developed models in two application scenarios which are of particular interest in computer vision: human bodies and natural scenes. In the practical sections of this work we discuss both areas from different angles and show applications of our models to human pose, motion, and segmentation as well as object categorization and localization. Here, we benefit from the availability of modern datasets of unprecedented size and diversity. Comparisons to traditional approaches and state-of-the-art research on the basis of well-established evaluation criteria allows the objective assessment of our contributions.

pdf [BibTex]

2014


Advanced Structured Prediction
Advanced Structured Prediction

Nowozin, S., Gehler, P. V., Jancsary, J., Lampert, C. H.

Advanced Structured Prediction, pages: 432, Neural Information Processing Series, MIT Press, November 2014 (book)

Abstract
The goal of structured prediction is to build machine learning models that predict relational information that itself has structure, such as being composed of multiple interrelated parts. These models, which reflect prior knowledge, task-specific relations, and constraints, are used in fields including computer vision, speech recognition, natural language processing, and computational biology. They can carry out such tasks as predicting a natural language sentence, or segmenting an image into meaningful components. These models are expressive and powerful, but exact computation is often intractable. A broad research effort in recent years has aimed at designing structured prediction models and approximate inference and learning procedures that are computationally efficient. This volume offers an overview of this recent research in order to make the work accessible to a broader research community. The chapters, by leading researchers in the field, cover a range of topics, including research trends, the linear programming relaxation approach, innovations in probabilistic modeling, recent theoretical progress, and resource-aware learning.

publisher link (url) [BibTex]

2014

publisher link (url) [BibTex]


Modeling the Human Body in 3D: Data Registration and Human Shape Representation
Modeling the Human Body in 3D: Data Registration and Human Shape Representation

Tsoli, A.

Brown University, Department of Computer Science, May 2014 (phdthesis)

pdf [BibTex]

pdf [BibTex]


Human Pose Estimation from Video and Inertial Sensors
Human Pose Estimation from Video and Inertial Sensors

Pons-Moll, G.

Ph.D Thesis, -, 2014 (book)

Abstract
The analysis and understanding of human movement is central to many applications such as sports science, medical diagnosis and movie production. The ability to automatically monitor human activity in security sensitive areas such as airports, lobbies or borders is of great practical importance. Furthermore, automatic pose estimation from images leverages the processing and understanding of massive digital libraries available on the Internet. We build upon a model based approach where the human shape is modelled with a surface mesh and the motion is parametrized by a kinematic chain. We then seek for the pose of the model that best explains the available observations coming from different sensors. In a first scenario, we consider a calibrated mult-iview setup in an indoor studio. To obtain very accurate results, we propose a novel tracker that combines information coming from video and a small set of Inertial Measurement Units (IMUs). We do so by locally optimizing a joint energy consisting of a term that measures the likelihood of the video data and a term for the IMU data. This is the first work to successfully combine video and IMUs information for full body pose estimation. When compared to commercial marker based systems the proposed solution is more cost efficient and less intrusive for the user. In a second scenario, we relax the assumption of an indoor studio and we tackle outdoor scenes with background clutter, illumination changes, large recording volumes and difficult motions of people interacting with objects. Again, we combine information from video and IMUs. Here we employ a particle based optimization approach that allows us to be more robust to tracking failures. To satisfy the orientation constraints imposed by the IMUs, we derive an analytic Inverse Kinematics (IK) procedure to sample from the manifold of valid poses. The generated hypothesis come from a lower dimensional manifold and therefore the computational cost can be reduced. Experiments on challenging sequences suggest the proposed tracker can be applied to capture in outdoor scenarios. Furthermore, the proposed IK sampling procedure can be used to integrate any kind of constraints derived from the environment. Finally, we consider the most challenging possible scenario: pose estimation of monocular images. Here, we argue that estimating the pose to the degree of accuracy as in an engineered environment is too ambitious with the current technology. Therefore, we propose to extract meaningful semantic information about the pose directly from image features in a discriminative fashion. In particular, we introduce posebits which are semantic pose descriptors about the geometric relationships between parts in the body. The experiments show that the intermediate step of inferring posebits from images can improve pose estimation from monocular imagery. Furthermore, posebits can be very useful as input feature for many computer vision algorithms.

pdf [BibTex]


Simulated Annealing
Simulated Annealing

Gall, J.

In Encyclopedia of Computer Vision, pages: 737-741, 0, (Editors: Ikeuchi, K. ), Springer Verlag, 2014, to appear (inbook)

[BibTex]

[BibTex]