TechTalks from event: CVPR 2014 Oral Talks

Orals 1C : Statistical Methods & Learning I

  • Anytime Recognition of Objects and Scenes Authors: Sergey Karayev, Mario Fritz, Trevor Darrell
    Humans are capable of perceiving a scene at a glance, and obtain deeper understanding with additional time. Similarly, visual recognition deployments should be robust to varying computational budgets. Such situations require Anytime recognition ability, which is rarely considered in computer vision research. We present a method for learning dynamic policies to optimize Anytime performance in visual architectures. Our model sequentially orders feature computation and performs subsequent classification. Crucially, decisions are made at test time and depend on observed data and intermediate results. We show the applicability of this system to standard problems in scene and object recognition. On suitable datasets, we can incorporate a semantic back-off strategy that gives maximally specific predictions for a desired level of accuracy; this provides a new view on the time course of human visual perception.
  • Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation Authors: Ross Girshick, Jeff Donahue, Trevor Darrell, Jitendra Malik
    Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012---achieving a mAP of 53.3%. Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also present experiments that provide insight into what the network learns, revealing a rich hierarchy of image features. Source code for the complete system is available at http://www.cs.berkeley.edu/~rbg/rcnn.
  • Optimal Decisions from Probabilistic Models: The Intersection-over-Union Case Authors: Sebastian Nowozin
    A probabilistic model allows us to reason about the world and make statistically optimal decisions using Bayesian decision theory. However, in practice the intractability of the decision problem forces us to adopt simplistic loss functions such as the 0/1 loss or Hamming loss and as result we make poor decisions through MAP estimates or through low-order marginal statistics. In this work we investigate optimal decision making for more realistic loss functions. Specifically we consider the popular intersection-over-union (IoU) score used in image segmentation benchmarks and show that it results in a hard combinatorial decision problem. To make this problem tractable we propose a statistical approximation to the objective function, as well as an approximate algorithm based on parametric linear programming. We apply the algorithm on three benchmark datasets and obtain improved intersection-over-union scores compared to maximum-posterior-marginal decisions. Our work points out the difficulties of using realistic loss functions with probabilistic computer vision models.
  • Covariance Trees for 2D and 3D Processing Authors: Thierry Guillemot, Andrés Almansa, Tamy Boubekeur
    Gaussian Mixture Models have become one of the major tools in modern statistical image processing, and allowed performance breakthroughs in patch-based image denoising and restoration problems. Nevertheless, their adoption level was kept relatively low because of the computational cost associated to learning such models on large image databases. This work provides a flexible and generic tool for dealing with such models without the computational penalty or parameter tuning difficulties associated to a na��ve implementation of GMM-based image restoration tasks. It does so by organising the data manifold in a hirerachical multiscale structure (the Covariance Tree) that can be queried at various scale levels around any point in feature-space. We start by explaining how to construct a Covariance Tree from a subset of the input data, how to enrich its statistics from a larger set in a streaming process, and how to query it efficiently, at any scale. We then demonstrate its usefulness on several applications, including non-local image filtering, data-driven denoising, reconstruction from random samples and surface modeling from unorganized 3D points sets.
  • Hierarchical Subquery Evaluation for Active Learning on a Graph Authors: Oisin Mac Aodha, Neill D.F. Campbell, Jan Kautz, Gabriel J. Brostow
    To train good supervised and semi-supervised object classifiers, it is critical that we not waste the time of the human experts who are providing the training labels. Existing active learning strategies can have uneven performance, being efficient on some datasets but wasteful on others, or inconsistent just between runs on the same dataset. We propose perplexity based graph construction and a new hierarchical subquery evaluation algorithm to combat this variability, and to release the potential of Expected Error Reduction. Under some specific circumstances, Expected Error Reduction has been one of the strongest-performing informativeness criteria for active learning. Until now, it has also been prohibitively costly to compute for sizeable datasets. We demonstrate our highly practical algorithm, comparing it to other active learning measures on classification datasets that vary in sparsity, dimensionality, and size. Our algorithm is consistent over multiple runs and achieves high accuracy, while querying the human expert for labels at a frequency that matches their desired time budget.