Daniel Tarlow

Learning Articulated Structure and Motion

Humans demonstrate a remarkable ability to parse complicated motion sequences into their constituent structures and motions. We investigate this problem, attempting to learn the structure of one or more articulated objects, given a time series of two-dimensional feature positions. We model the observed sequence in terms of “stick figure” objects, under the assumption that the relative joint angles between sticks can change over time, but their lengths and connectivities are fixed. The problem is formulated as a single probabilistic model that includes multiple sub-components: associating the features with particular sticks, determining the proper number of sticks, and finding which sticks are physically joined. We test the algorithm on challenging datasets of 2D projections of optical human motion capture and feature trajectories from real videos.

Exploring Compositional High Order Pattern Potentials for Structured Output Learning

When modeling structured outputs such as image segmentations, prediction can be improved by accurately modeling structure present in the labels. A key challenge is developing tractable models that are able to capture complex high level structure like shape. In this work, we study the learning of a general class of pattern-like high order potential, which we call Compositional High Order Pattern Potentials (CHOPPs). We show that CHOPPs include the linear deviation pattern potentials of Rother et al. [26] and also Restricted Boltzmann Machines (RBMs), we also establish the near equivalence of these two models. Experimentally, we show that performance is affected significantly by the degree of variability present in the datasets, and we define a quantitative variability measure to aid in studying this. We then improve CHOPPs performance in high variability datasets with two primary contributions: (a) developing a loss-sensitive joint learning procedure, so that internal pattern parameters can be learned in conjunction with other model potentials to minimize expected loss, and (b) learning an image-dependent mapping that encourages or inhibits patterns depending on image features. We also explore varying how multiple patterns are composed, and learning convolutional patterns. Quantitative results on challenging highly variable datasets show that the joint learning and image-dependent high order potentials can improve performance.

Fits Like a Glove: Rapid and Reliable Hand Shape Personalization

We present a fast, practical method for personalizing a hand shape basis to an individual users detailed hand shape using only a small set of depth images. To achieve this, we minimize an energy based on a sum of render-and-compare cost functions called the golden energy. However, this energy is only piecewise continuous, due to pixels crossing occlusion boundaries, and is therefore not obviously amenable to efficient gradient-based optimization. A key insight is that the energy is the combination of a smooth low-frequency function with a high-frequency, low-amplitude, piecewisecontinuous function. A central finite difference approximation with a suitable step size can therefore jump over the discontinuities to obtain a good approximation to the energys low-frequency behavior, allowing efficient gradient-based optimization. Experimental results quantitatively demonstrate for the first time that detailed personalized models improve the accuracy of hand tracking and achieve competitive results in both tracking and model registration.

Empirical Minimum Bayes Risk Prediction: How to Extract an Extra Few % Performance from Vision Models with Just Three More Parameters

When building vision systems that predict structured objects such as image segmentations or human poses, a crucial concern is performance under task-specific evaluation measures (e.g. Jaccard Index or Average Precision). An ongoing research challenge is to optimize predictions so as to maximize performance on such complex measures. In this work, we present a simple meta-algorithm that is surprisingly effective -- Empirical Min Bayes Risk. EMBR takes as input a pre-trained model that would normally be the final product and learns three additional parameters so as to optimize performance on the complex high-order task-specific measure. We demonstrate EMBR in several domains, taking existing state-of-the-art algorithms and improving performance up to ~7%, simply with three extra parameters.

Revisiting uncertainty in graph cut solutions

Graph cuts is a popular algorithm for finding the MAP assignment of many large-scale graphical models that are common in computer vision. While graph cuts is powerful, it does not provide information about the marginal probabilities associated with the solution it finds. To assess uncertainty, we are forced to fall back on less efficient and inexact inference algorithms such as loopy belief propagation, or use less principled surrogate representations of uncertainty such as the min-marginal approach of Kohli & Torr [8]. In this work, we give new justification for using min-marginals to compute the uncertainty in conditional random fields, framing the min-marginal outputs as exact marginals under a specially-chosen generative probabilistic model. We leverage this view to learn properly calibrated marginal probabilities as the result of straightforward maximization of the training likelihood, showing that the necessary subgradients can be computed efficiently using dynamic graph cut operations. We also show how this approach can be extended to compute multi-label marginal distributions, where again dynamic graph cuts enable efficient marginal inference and maximum likelihood learning. We demonstrate empirically that - after proper training - uncertainties based on min-marginals provide better-calibrated probabilities than baselines and that these distributions can be exploited in a decision-theoretic way for improved segmentation in low-level vision.

Optimizing Expected Intersection-Over-Union with Candidate-Constrained CRFs

We study the question of how to make loss-aware predictions in image segmentation settings where the evaluation function is the Intersection-over-Union (IoU) measure that is used widely in evaluating image segmentation systems. Currently, there are two dominant approaches: the first approximates the Expected-IoU (EIoU) score as Expected-Intersection-over-Expected-Union (EIoEU), and the second approach is to compute exact EIoU but only over a small set of high-quality candidate solutions. We begin by asking which approach we should favor for two typical image segmentation tasks. Studying this question leads to two new methods that draw ideas from both existing approaches. Our new methods use the EIoEU approximation paired with high quality candidate solutions. Experimentally we show that our new approaches lead to improved performance on both image segmentation tasks.

Probabilistic Programs as Spreadsheet Queries

We describe the design, semantics, and implementation of a probabilistic programming language where programs are spreadsheet queries. Given an input database consisting of tables held in a spreadsheet, a query constructs a probabilistic model conditioned by the spreadsheet data, and returns an output database determined by inference. This work extends probabilistic programming systems in three novel aspects: (1) embedding in spreadsheets, (2) dependently typed functions, and (3) typed distinction between random and query variables. It empowers users with knowledge of statistical modelling to do inference simply by editing textual annotations within their spreadsheets, with no other coding.

Learning Shape Analysis

We present a data-driven verification framework to automatically prove memory safety of heap-manipulating programs. Our core contribution is a novel statistical machine learning technique that maps observed program states to (possibly disjunctive) separation logic formulas describing the invariant shape of (possibly nested) data structures at relevant program locations. We then attempt to verify these predictions using a program verifier, where counterexamples to a predicted invariant are used as additional input to the shape predictor in a refinement loop. We have implemented our techniques in Locust, an extension of the GRASShopper verification tool. Locust is able to automatically prove memory safety of implementations of classical heap-manipulating programs such as insertionsort, quicksort and traversals of nested data structures.

Empirical Minimum Bayes Risk Prediction

When building vision systems that predict structured objects such as image segmentations or human poses, a crucial concern is performance under task-specific evaluation measures (e.g., Jaccard Index or Average Precision). An ongoing research challenge is to optimize predictions so as to maximize performance on such complex measures. In this work, we present a simple meta-algorithm that is surprisingly effective – Empirical Min Bayes Risk. EMBR takes as input a pre-trained model that would normally be the final product and learns three additional parameters so as to optimize performance on the complex instance-level high-order task-specific measure. We demonstrate EMBR in several domains, taking existing state-of-the-art algorithms and improving performance up to 8 percent, simply by learning three extra parameters. Our code is publicly available and the results presented in this paper can be replicated from our code-release.