CVPR 2016 Papers

We revisit the idea of brain damage, i.e. the pruning of the coefficients of a neural network, and suggest how brain damage can be modified and used to speedup convolutional layers in ConvNets. The approach uses the fact that many efficient implementations reduce generalized convolutions to matrix multiplications. The suggested brain damage process prunes the convolutional kernel tensor in a group-wise fashion. After such pruning, convolutions can be reduced to multiplications of thinned dense matrices, which leads to speedup. We investigate different ways to add group-wise prunning to the learning process, and show that several-fold speedups of convolutional layers can be attained using group-sparsity regularizers. Our approach can adjust the shapes of the receptive fields in the convolutional layers, and even prune excessive feature maps from ConvNets, all in a data-driven way.

The SVM+ approach has shown excellent performance in visual recognition tasks for exploiting privileged information in the training data. In this paper, we propose two efficient algorithms for solving the linear and kernel SVM+, respectively. For linear SVM+, we absorb the bias term into the weight vector, and formulate a new optimization problem with simpler constraints in the dual form. Then, we develop an efficient dual coordinate descent algorithm to solve the new optimization problem. For kernel SVM+, we further apply the l2-loss, which leads to a simpler optimization problem in the dual form with only half of dual variables when compared with the dual form of the original SVM+ method. More interestingly, we show that our new dual problem can be efficiently solved by using the SMO algorithm of the one-class SVM problem. Comprehensive experiments on three datasets clearly demonstrate that our proposed algorithms achieve significant speed-up than the state-of-the-art solvers for linear and kernel SVM+.

Deep convolutional neural networks take GPU-days of computation to train on large data sets. Pedestrian detection for self driving cars requires very low latency. Image recognition for mobile phones is constrained by limited processing resources. The success of convolutional neural networks in these situations is limited by how fast we can compute them. Conventional FFT based convolution is fast for large filters, but state of the art convolutional neural networks use small, 3x3 filters. We introduce a new class of fast algorithms for convolutional neural networks using Winograd's minimal filtering algorithms. The algorithms compute minimal complexity convolution over small tiles, which makes them fast with small filters and small batch sizes. We benchmark a GPU implementation of our algorithm with the VGG network and show state of the art throughput at batch sizes from 1 to 64.

We propose a novel approach for learning features from weakly-supervised data by joint ranking and classification. In order to exploit data with weak labels, we jointly train a feature extraction network with a ranking loss and a classification network with a cross-entropy loss. We obtain high-quality compact discriminative features with few parameters, learned on relatively small datasets without additional annotations. This enables us to tackle tasks with specialized images not very similar to the more generic ones in existing fully-supervised datasets. We show that the resulting features in combination with a linear classifier surpass the state-of-the-art on the Hipster Wars dataset despite using features only 0.3% of the size. Our proposed features significantly outperform those obtained from networks trained on ImageNet, despite being 32 times smaller (128 single-precision floats), trained on noisy and weakly-labeled data, and using only 1.5% of the number of parameters.

Finetuning from a pretrained deep model is found to yield state-of-the-art performance for many vision tasks. This paper investigates many factors that influence the performance in finetuning for object detection. There is a long-tailed distribution of sample numbers for classes in object detection. Our analysis and empirical results show that classes with more samples have higher impact on the feature learning. And it is better to make the sample number more uniform across classes. Generic object detection can be considered as multiple equally important tasks. Detection of each class is a task. These classes/tasks have their individuality in discriminative visual appearance representation. Taking this individuality into account, we cluster objects into visually similar class groups and learn deep representations for these groups separately. A hierarchical feature learning scheme is proposed. In this scheme, the knowledge from the group with large number of classes is transferred for learning features in its sub-groups. Finetuned on the GoogLeNet model, experimental results show 4.7% absolute mAP improvement of our approach on the ImageNet object detection dataset without increasing much computational cost at the testing stage.

Previous studies on facial expression analysis have been focused on recognizing basic expression categories. There is limited amount of work on the continuous expression intensity estimation, which is important for detecting and tracking emotion change. Part of the reason is the lack of labeled data with annotated expression intensity since expression intensity annotation requires expertise and is time consuming. In this work, we treat the expression intensity estimation as a regression problem. By taking advantage of the natural onset-apex-offset evolution pattern of facial expression, the proposed method can handle different amounts of annotations to perform frame-level expression intensity estimation. In fully supervised case, all the frames are provided with intensity annotations. In weakly supervised case, only the annotations of selected key frames are used. While in unsupervised case, expression intensity can be estimated without any annotations. An efficient optimization algorithm based on Alternating Direction Method of Multipliers (ADMM) is developed for solving the optimization problem associated with parameter learning. We demonstrate the effectiveness of proposed method by comparing it against both fully supervised and unsupervised approaches on benchmark facial expression datasets.

We present a novel approach for real-time facial reenactment of a monocular target video sequence (e.g., Youtube video). The source sequence is also a monocular video stream, captured live with a commodity webcam. Our goal is to animate the facial expressions of the target video by a source actor and re-render the manipulated output video in a photo-realistic fashion. To this end, we first address the under-constrained problem of facial identity recovery from monocular video by non-rigid model-based bundling. At run time, we track facial expressions of both source and target video using a dense photometric consistency measure. Reenactment is then achieved by fast and efficient deformation transfer between source and target. The mouth interior that best matches the re-targeted expression is retrieved from the target sequence and warped to produce an accurate fit. Finally, we convincingly re-render the synthesized target face on top of the corresponding video stream such that it seamlessly blends with the real-world illumination. We demonstrate our method in a live setup, where Youtube videos are reenacted in real time.

Face alignment, which fits a face model to an image and extracts the semantic meanings of facial pixels, has been an important topic in CV community. However, most algorithms are designed for faces in small to medium poses (below 45 degree), lacking the ability to align faces in large-pose up to 90 degree. The challenges are three-fold: Firstly, the commonly used landmark-based face model assumes that all the landmarks are visible and is therefore not suitable for profile views. Secondly, the face appearance varies more dramatically in large poses, ranging from frontal view to profile view. Thirdly, labelling landmarks in large poses is an extremely challenging work since the invisible landmarks have to be guessed. In this paper, we propose a solution to the three problems in an new alignment framework, called 3D Dense Face Alignment (3DDFA), in which a dense 3D face model is fitted to the image via convolutional neutral network (CNN). We also propose a method to synthesize large-scale training samples in profile views to solve the third problem of data labelling. Experiments on the challenging AFLW database show that our approach achieves significant improvements over state-of-the-art methods.

We present a new method for approximate nearest neighbour search on large datasets of high dimensional feature vectors, such as SIFT or GIST descriptors. Our approach constructs a directed graph that can be efficiently explored for nearest neighbour queries. Each vertex in this graph represents a feature vector from the dataset being searched. The directed edges are computed by exploiting the fact that, for these datasets, the intrinsic dimensionality of the local manifold-like structure formed by the elements of the dataset is significantly lower than the embedding space. We also provide an efficient search algorithm that uses this graph to rapidly find the nearest neighbour to a query with high probability. We show how the method can be adapted to give a strong guarantee of 100% recall where the query is within a threshold distance of its nearest neighbour. We demonstrate that our method is significantly more efficient than existing state of the art methods. In particular, our GPU implementation can deliver 90% recall for queries on a data set of 1 million SIFT descriptors at a rate of over 1.2 million queries per second on a Titan X. Finally we also demonstrate how our method scales to datasets of 5M and 20M entries.

From scientific research to commercial applications, eye tracking is an important tool across many domains. Despite its range of applications, eye tracking has yet to become a pervasive technology. We believe that we can put the power of eye tracking in everyone's palm by building eye tracking software that works on commodity hardware such as mobile phones and tablets, without the need for additional sensors or devices. We tackle this problem by introducing GazeCapture, the first large-scale dataset for eye tracking, containing data from over 1450 people consisting of almost 2:5M frames. Using GazeCapture, we train iTracker, a convolutional neural network for eye tracking, which achieves a significant reduction in error over previous approaches while running in real time (10-15fps) on a modern mobile device. Our model achieves a prediction error of 1.71cm and 2.53cm without calibration on mobile phones and tablets respectively. With calibration, this is reduced to 1.34cm and 2.12cm. Further, we demonstrate that the features learned by iTracker generalize well to other datasets, achieving state-of-the-art results. The code, data, and models are available at http://gazecapture.csail.mit.edu.

Conventional scanning and multiplexing techniques for hyperspectral imaging suffer from limited temporal and/or spatial resolution. To resolve this issue, coding techniques are becoming increasingly popular in developing snapshot systems for high-resolution hyperspectral imaging. For such systems, it is a critical task to accurately restore the 3D hyperspectral image from its corresponding coded 2D image. In this paper, we propose an effective method for coded hyperspectral image restoration, which exploits extensive structure sparsity in the hyperspectral image. Specifically, we simultaneously explore spectral and spatial correlation via low-rank regularizations, and formulate the restoration problem into a variational optimization model, which can be solved via an iterative numerical algorithm. Experimental results using both synthetic data and real images show that the proposed method can significantly outperform the state-of-the-art methods on several popular coding-based hyperspectral imaging systems.

Convolutional neural networks (CNNs) have shown great performance as general feature representations for object recognition applications. However, for multi-label images that contain multiple objects from different categories, scales and locations, global CNN features are not optimal. In this paper, we incorporate local information to enhance the feature discriminative power. In particular, we first extract object proposals from each image. With each image treated as a bag and object proposals extracted from it treated as instances, we transform the multi-label recognition problem into a multi-class multi-instance learning problem. Then, in addition to extracting the typical CNN feature representation from each proposal, we propose to make use of ground-truth bounding box annotations (strong labels) to add another level of local information by using nearest-neighbor relationships of local regions to form a multi-view pipeline. The proposed multi-view multi-instance framework utilizes both weak and strong labels effectively, and more importantly it has the generalization ability to even boost the performance of unseen categories by partial strong labels from other categories. Our framework is extensively compared with state-of-the-art hand-crafted feature based methods and CNN based methods on two multi-label benchmark datasets. The experimental results validate the discriminative power and the generalization ability of the proposed framework. With strong labels, our framework is able to achieve state-of-the-art results in both datasets.

In this paper, we investigate two new strategies to detect objects accurately and efficiently using deep convolutional neural network: 1) scale-dependent pooling and 2) layer-wise cascaded rejection classifiers. The scale-dependent pooling (SDP) improves detection accuracy by exploiting appropriate convolutional features depending on the scale of candidate object proposals. The cascaded rejection classifiers (CRC) effectively utilize convolutional features and eliminate negative object proposals in a cascaded manner, which greatly speeds up the detection while maintaining high accuracy. In combination of the two, our method achieves significantly better accuracy compared to other state-of-the-arts in three challenging datasets, PASCAL object detection challenge, KITTI object detection benchmark and newly collected Inner-city dataset, while being more efficient.

Top-down saliency detection is a knowledge-driven search task. While some previous methods aim to learn this "knowledge" from category-specific data, others transfer existing annotations in a large dataset through appearance matching. In contrast, we propose in this paper a locate-by-exemplar strategy. This approach is challenging, as we only use a few exemplars (up to 4) and the appearances among the query object and the exemplars can be very different. To address it, we design a two-stage deep model to learn the intra-class association between the exemplars and query objects. The first stage is for learning object-to-object association, and the second stage is to learn background discrimination. Extensive experimental evaluations show that the proposed method outperforms different baselines and the category-specific models. In addition, we explore the influence of exemplar properties, in terms of exemplar number and quality. Furthermore, we show that the learned model is a universal model and offers great generalization to unseen objects.

When creating a photo album of an event, people typically select a few important images to keep or share. There is some consistency in the process of choosing the important images, and discarding the unimportant ones. Modeling this selection process will assist automatic photo selection and album summarization. In this paper, we show that the selection of important images is consistent among different viewers, and that this selection process is related to the event type of the album. We introduce the concept of event-specific image importance. We collected a new event album dataset with human annotation of the relative image importance with each event album. We also propose a Convolutional Neural Network (CNN) based method to predict the image importance score of a given event album, using a novel rank loss function and a progressive training scheme. Results demonstrate that our method significantly outperforms various baseline methods.

Conventional algorithms for sparse signal recovery and sparse representation rely on l1-norm regularized variational methods. However, when applied to the reconstruction of sparse images, i.e., images where only a few pixels are non-zero, simple l1-norm-based methods ignore poten- tial correlations in the support between adjacent pixels. In a number of applications, one is interested in images that are not only sparse, but also have a support with smooth (or contiguous) boundaries. Existing algorithms that take into account such a support structure mostly rely on non-convex methods and--as a consequence--do not scale well to high-dimensional problems and/or do not converge to global optima. In this paper, we explore the use of new block l1-norm regularizers, which enforce image sparsity while simultaneously promoting smooth support structure. By exploiting the convexity of our regularizers, we develop new computationally-efficient recovery algorithms that guarantee global optimality. We demonstrate the efficacy of our regularizers on a variety of imaging tasks including compressive image recovery, image restoration, and robust PCA.

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.

Most existing dictionary learning algorithms consider a linear sparse model, which often cannot effectively characterize the nonlinear properties present in many types of visual data, e.g. dynamic texture (DT). Such nonlinear properties can be exploited by the so-called kernel sparse coding. This paper proposed an equiangular kernel dictionary learning method with optimal mutual coherence to exploit the nonlinear sparsity of high-dimensional visual data. Two main issues are addressed in the proposed method: (1) coding stability for redundant dictionary of infinite-dimensional space; and (2) computational efficiency for computing kernel matrix of training samples of high-dimensional data. The proposed kernel sparse coding method is applied to dynamic texture analysis with both local DT pattern extraction and global DT pattern characterization. The experimental results showed its performance gain over existing methods.

Most saliency estimation methods aim to explicitly model low-level conspicuity cues such as edges or blobs and may additionally incorporate top-down cues using face or text detection. Data-driven methods for training saliency models using eye-fixation data are increasingly popular, particularly with the introduction of large-scale datasets and deep architectures. However, current methods in this latter paradigm use loss functions designed for classification or regression tasks whereas saliency estimation is evaluated on topographical maps. In this work, we introduce a new saliency map model which formulates a map as a generalized Bernoulli distribution. We then train a deep architecture to predict such maps using novel loss functions which pair the softmax activation function with measures designed to compute distances between probability distributions. We show in extensive experiments the effectiveness of such loss functions over standard ones on four public benchmark datasets, and demonstrate improved performance over state-of-the-art saliency methods.