CVPR 2016 Papers

This paper presents a novel scene text detection algorithm, Canny Text Detector, which takes advantage of the similarity between image edge and text for effective text localization with improved recall rate. As closely related edge pixels construct the structural information of an object, we observe that cohesive characters compose a meaningful word/sentence sharing similar properties such as spatial location, size, color, and stroke width regardless of language. However, prevalent scene text detection approaches have not fully utilized such similarity, but mostly rely on the characters classified with high confidence, leading to low recall rate. By exploiting the similarity, our approach can quickly and robustly localize a variety of texts. Inspired by the original Canny edge detector, our algorithm makes use of double threshold and hysteresis tracking to detect texts of low confidence. Experimental results on public datasets demonstrate that our algorithm outperforms the state-of-the-art scene text detection methods in terms of detection rate.

Camera calibration directly from image sequences of a pedestrian without using any calibration object is a really challenging task and should be well solved in computer vision, especially in visual surveillance. In this paper, we propose a novel camera calibration method based on recovering the three orthogonal vanishing points (TOVPs), just using an image sequence of a pedestrian walking in a straight line, without any assumption of scenes or motions, e.g., control points with known 3D coordinates, parallel or perpendicular lines, non-natural or pre-designed special human motions, as often necessary in previous methods. The traces of shoes of a pedestrian carry more rich and easily detectable metric information than all other body parts in the periodic motion of a pedestrian, but such information is usually overlooked by previous work. In this paper, we employ the images of the toes of the shoes on the ground plane to determine the vanishing point corresponding to the walking direction, and then utilize harmonic conjugate properties in projective geometry to recover the vanishing point corresponding to the perpendicular direction of the walking direction in the horizontal plane and the vanishing point corresponding to the vertical direction. After recovering all of the TOVPs, the intrinsic and extrinsic parameters of the camera can be determined. Experiments on various scenes and viewing angles prove the feasibility and accuracy of the proposed method.

Computing the epipolar geometry between cameras with very different viewpoints is often problematic as matching points are hard to find. In these cases, it has been proposed to use information from dynamic objects in the scene for suggesting point and line correspondences. We propose a speed up of about two orders of magnitude, as well as an increase in robustness and accuracy, to methods computing epipolar geometry from dynamic silhouettes based on a new temporal signature, motion barcode for lines. This is a binary temporal sequence for lines, indicating for each frame the existence of at least one foreground pixel on that line. The motion barcodes of two corresponding epipolar lines are very similar so the search for corresponding epipolar lines can be limited to lines having similar barcodes leading to increased speed, accuracy, and robustness in computing the epipolar geometry.

Object detection is a fundamental problem in image understanding. One popular solution is the R-CNN framework and its fast versions. They decompose the object detection problem into two cascaded easier tasks: 1) generating object proposals from images, 2) classifying proposals into various object categories. Despite that we are handling with two relatively easier tasks, they are not solved perfectly and there's still room for improvement. In this paper, we push the "divide and conquer" solution even further by dividing each task into two sub-tasks. We call the proposed method "CRAFT" (Cascade Region-proposal-network And FasT-rcnn), which tackles each task with a carefully designed network cascade. We show that the cascade structure helps in both tasks: in proposal generation, it provides more compact and better localized object proposals; in object classification, it reduces false positives (mainly between ambiguous categories) by capturing both inter- and intra-category variances. CRAFT achieves consistent and considerable improvement over the state-of-the-art on object detection benchmarks like PASCAL VOC 07/12 and ILSVRC.

We propose a novel Coupled Projection multi-task Met- ric Learning (CP-mtML) method for large scale face re- trieval. In contrast to previous works which were limited to low dimensional features and small datasets, the proposed method scales to large datasets with high dimensional face descriptors. It utilises pairwise (dis-)similarity constraints as supervision and hence does not require exhaustive class annotation for every training image. While, traditionally, multi-task learning methods have been validated on same dataset but different tasks, we work on the more chal- lenging setting with heterogeneous datasets and different tasks. We show empirical validation on multiple face im- age datasets of different facial traits, e.g. identity, age and expression. We use classic Local Binary Pattern (LBP) de- scriptors along with the recent Deep Convolutional Neural Network (CNN) features. The experiments clearly demon- strate the scalability and improved performance of the pro- posed method on the tasks of identity and age based face image retrieval compared to competitive existing methods, on the standard datasets and with the presence of a million distractor face images.

While deep convolutional neural networks (CNNs) have shown a great success in single-label image classification, it is important to note that most real world images contain multiple labels, which could correspond to different objects, scenes, actions and attributes in an image. Traditional approaches to multi-label image classification learn independent classifiers for each category and employ ranking or thresholding on the classification results. These techniques, although working well, fail to explicitly exploit the label dependencies in an image. In this paper, we utilize recurrent neural networks (RNNs) to address this problem. Combined with CNNs, the proposed CNN-RNN framework learns a joint image-label embedding to characterize the semantic label dependency as well as the image-label relevance, and it can be trained end-to-end from scratch to integrate both information in an unified framework. Experimental results on public benchmark datasets demonstrate that the proposed architecture achieves better performance than the state-of-the-art multi-label classification models.

Given an image of a handwritten word, a CNN is employed to estimate its n-gram frequency profile, which is the set of n-grams contained in the word. Frequencies for unigrams, bigrams and trigrams are estimated for the entire word and for parts of it. Canonical Correlation Analysis is then used to match the estimated profile to the true profiles of all words in a large dictionary. The CNN that is used employs several novelties such as the use of multiple fully connected branches. Applied to all commonly used handwriting recognition benchmarks, our method outperforms, by a very large margin, all existing methods.

We are dealing with the problem of fine-grained vehicle make&model recognition and verification. Our contribution is showing that extracting additional data from the video stream - besides the vehicle image itself - and feeding it into the deep convolutional neural network boosts the recognition performance considerably. This additional information includes: 3D vehicle bounding box used for "unpacking" the vehicle image, its rasterized low-resolution shape, and information about the 3D vehicle orientation. Experiments show that adding such information decreases classification error by 26% (the accuracy is improved from 0.772 to 0.832) and boosts verification average precision by 208% (0.378 to 0.785) compared to baseline pure CNN without any input modifications. Also, the pure baseline CNN outperforms the recent state of the art solution by 0.081. We provide an annotated set "BoxCars" of surveillance vehicle images augmented by various automatically extracted auxiliary information. Our approach and the dataset can considerably improve the performance of traffic surveillance systems.

Convolutional neural nets (CNNs) have demonstrated remarkable performance in recent history. Such approaches tend to work in a "unidirectional" bottom-up feed-forward fashion. However, practical experience and biological evidence tells us that feedback plays a crucial role, particularly for detailed spatial understanding tasks. This work explores "bidirectional" architectures that also reason with top-down feedback: neural units are influenced by both lower and higher-level units. We do so by treating units as rectified latent variables in a quadratic energy function, which can be seen as a hierarchical Rectified Gaussian model (RGs). We show that RGs can be optimized with a quadratic program (QP), that can in turn be optimized with a recurrent neural network (with rectified linear units). This allows RGs to be trained with GPU-optimized gradient descent. From a theoretical perspective, RGs help establish a connection between CNNs and hierarchical probabilistic models. From a practical perspective, RGs are well suited for detailed spatial tasks that can benefit from top-down reasoning. We illustrate them on the challenging task of keypoint localization under occlusions, where local bottom-up evidence may be misleading. We demonstrate state-of-the-art results on challenging benchmarks.

Most deep architectures for image classification--even those that are trained to classify a large number of diverse categories--learn shared image representations with a single model. Intuitively, however, categories that are more similar should share more information than those that are very different. While hierarchical deep networks address this problem by learning separate features for subsets of related categories, current implementations require simplified models using fixed architectures specified via heuristic clustering methods. Instead, we propose Blockout, a method for regularization and model selection that simultaneously learns both the model architecture and parameters. A generalization of Dropout, our approach gives a novel parametrization of hierarchical architectures that allows for structure learning via back-propagation. To demonstrate its utility, we evaluate Blockout on the CIFAR and ImageNet datasets, demonstrating improved classification accuracy, better regularization performance, faster training, and the clear emergence of hierarchical network structures.

Blind image deconvolution is an ill-posed inverse problem which is often addressed through the application of appropriate prior. Although some priors are informative in general, many images do not strictly conform to this, leading to degraded performance in the kernel estimation. More critically, real images may be contaminated by nonuniform noise such as saturation and outliers.Methods for removing specific image areas based on some priors have been proposed, but they operate either manually or by defining fixed criteria. We show here that a subset of the image gradients are adequate to estimate the blur kernel robustly, no matter the gradient image is sparse or not. We thus introduce a gradient activation method to automatically select a subset of gradients of the latent image in a cutting-plane-based optimization scheme for kernel estimation. No extra assumption is used in our model, which greatly improves the accuracy and flexibility. More importantly, the proposed method affords great convenience for handling noise and outliers. Experiments on both synthetic data and real-world images demonstrate the effectiveness and robustness of the proposed method in comparison with the state-of-the-art methods.

We present a simple and effective blind image deblurring method based on the dark channel prior. Our work is inspired by the interesting observation that the dark channel of blurred images is less sparse. While most image patches in the clean image contain some dark pixels, these pixels are not dark when averaged with neighboring high-intensity pixels during the blur process.Our analysis shows that this change in the sparsity of the dark channel is an inherent property of the blur process, both theoretically and empirically. This change in the sparsity of the dark channel is an inherent property of the blur process, which we both prove mathematically and validate using training data. Therefore, enforcing the sparsity of the dark channel helps blind deblurring on various scenarios, including natural, face, text, and low-illumination images. However, sparsity of the dark channel introduces a non-convex non-linear optimization problem. We introduce a linear approximation of the min operator to compute the dark channel. Our look-up-table-based method converges fast in practice and can be directly extended to non-uniform deblurring. Extensive experiments show that our method achieves state-of-the-art results on deblurring natural images and compares favorably methods that are well-engineered for specific scenarios.

In this work, we propose a novel approach to video segmentation that operates in bilateral space. We design a new energy on the vertices of a regularly sampled spatio-temporal bilateral grid, which can be solved efficiently using a standard graph cut label assignment. Using a bilateral formulation, the energy that we minimize implicitly approximates long-range, spatio-temporal connections between pixels while still containing only a small number of variables and only local graph edges. We compare to a number of recent methods, and show that our approach achieves state-of-the-art results on multiple benchmarks in a fraction of the runtime. Furthermore, our method scales linearly with image size, allowing for interactive feedback on real-world high resolution video.

Most tracking-by-detection methods employ a local search window around the predicted object location in the current frame assuming the previous location is accurate, the trajectory is smooth, and the computational capacity permits a search radius that can accommodate the maximum speed yet small enough to reduce mismatches. These, however, may not be valid always, in particular for fast and irregularly moving objects. Here, we present an object tracker that is not limited to a local search window and has ability to probe efficiently the entire frame. Our method generates a small number of "high-quality" proposals by a novel instance-specific objectness measure and evaluates them against the object model that can be adopted from an existing tracking-by-detection approach as a core tracker. During the tracking process, we update the object model concentrating on hard false-positives supplied by the proposals, which help suppressing distractors caused by difficult background clutters, and learn how to re-rank proposals according to the object model. Since we reduce significantly the number of hypotheses the core tracker evaluates, we can use richer object descriptors and stronger detector. Our method outperforms most recent state-of-the-art trackers on popular tracking benchmarks, and provides improved robustness for fast moving objects as well as for ultra low-frame-rate videos.

In this paper, we present the first attempt to analyse differing levels of social involvement in free standing conversing groups (or the so-called F-formations) from static images. In addition, we enrich state-of-the-art F-formation modelling by learning a frustum of attention that accounts for the spatial context. That is, F-formation configurations vary with respect to the arrangement of furniture and the non-uniform crowdedness in the space during mingling scenarios. The majority of prior works have considered the labelling of conversing group as an objective task, requiring only a single annotator. However, we show that by embracing the subjectivity of social involvement, we not only generate a richer model of the social interactions in a scene but also significantly improve F-formation detection. We carry out extensive experimental validation of our proposed approach by collecting a novel set of multi-annotator labels of involvement on the publicly available Idiap Poster Data; the only multi-annotator labelled database of free standing conversing groups that is currently available.

Top-down saliency models produce a probability map that peaks at target locations specified by a task/goal such as object detection. They are usually trained in a supervised setting involving annotations of objects. We propose a weakly supervised top-down saliency framework using only binary labels that indicate the presence/absence of an object in an image. First, the probabilistic contribution of each image patch to the confidence of an ScSPM-based classifier produces a Reverse-ScSPM (R-ScSPM) saliency map. Neighborhood information is then incorporated through a contextual saliency map which is estimated using logistic regression learnt on patches having high R-ScSPM saliency. Both the saliency maps are combined to obtain the final saliency map. We evaluate the performance of the proposed weakly supervised top-down saliency and achieves comparable performance with fully supervised approaches. Experiments are carried out on 5 challenging datasets across 3 different applications.

This paper presents an algorithm coined BORDER (Bounding Oriented-Rectangle Descriptors for Enclosed Regions) for texture-less object recognition. By fusing a regional object encompassment concept with descriptor-based pipelines, we extend local-patches into scalable object-sized oriented rectangles for optimal object information encapsulation with minimal outliers. We correspondingly introduce a modified line-segment detection technique termed Linelets to stabilize keypoint repeatability in homogenous conditions. In addition, a unique sampling technique facilitates the incorporation of robust angle primitives to produce discriminative rotation-invariant descriptors. BORDER's high competence in object recognition particularly excels in homogenous conditions obtaining superior detection rates in the presence of high-clutter, occlusion and scale-rotation changes when compared with modern state-of-the-art texture-less object detectors such as BOLD and LINE2D on public texture-less object databases.

Cardiovascular disease (CVD) is the leading cause of mortality yet largely preventable, but the key to prevention is to identify at risk individuals before adverse events. For predicting individual CVD risk, carotid intima-media thickness (CIMT), a noninvasive ultrasound method, has proven to be valuable, offering several advantages over CT coronary artery calcium score. However, each CIMT examination includes several ultrasound videos, and interpreting each of these CIMT videos involves three operations: (1) select three end-diastolic ultrasound frames (EUF) in the video, (2) localize a region of interest (ROI) in each selected frame, and (3) trace the lumen-intima interface and the media-adventitia interface in each ROI to measure CIMT. These operations are tedious, laborious, and time consuming, a serious limitation that hinders the widespread utilization of CIMT in clinical practice. To overcome this limitation, this paper presents a new system to automate CIMT video interpretation. Our extensive experiments demonstrate that the suggested system significantly outperforms the state-of-the-art methods. The superior performance is attributable to our unified framework based on convolutional neural networks (CNNs) coupled with our informative image representation and effective post-processing of the CNN outputs, which are uniquely designed for each of the above three operations.

Attention based automatic image cropping aims at preserving the most visually important region in an image. A common task in this kind of method is to search for the smallest rectangle inside which the summed attention is maximized. We demonstrate that under appropriate formulations, this task can be achieved using efficient algorithms with low computational complexity. In a practically useful scenario where the aspect ratio of the cropping rectangle is given, the problem can be solved with a computational complexity linear to the number of image pixels. We also study the possibility of multiple rectangle cropping and a new model facilitating fully automated image cropping.