This paper introduces a method to generate duet dance videos from an input solo dancer’s video performance. Addressing this novel problem, our system tackles a set of sub-tasks, including dancer segmentation, camera motion handling, stage reconstruction and the intricate management of geometric constraints such as dancer scale preservation and dancers collision prevention. The proposed approach leverages existing methodologies and new solutions. Notably, we address collisions in 2D space–time directly, departing from traditional 3D approaches. We modify the initial location of the dancer to avoid long-time collisions globally, while also modulating the pace of the dance by deliberate slowing down or accelerating motion to avoid short collisions locally. Experimental results attest to the efficacy of our approach. The system not only successfully synthesizes engaging duet dance sequences but also upholds the authenticity of individual performances, as shown by a user study.

Consider the geo-localization task of finding the pose of a camera in a large 3D scene from a single image. Most existing CNN-based methods use as input textured images. We aim to experimentally explore whether texture and correlation between nearby images are necessary in a CNN-based solution for the geo-localization task. To do so, we consider lean images , textureless projections of a simple 3D model of a city. They only contain information related to the geometry of the scene viewed (edges, faces, and relative depth). The main contributions of this paper are: (i) to demonstrate the ability of CNNs to recover camera pose using lean images; and (ii) to provide insight into the role of geometry in the CNN learning process.

Automatic music transcription has primarily focused on transcribing audio to a symbolic music representation (e.g. MIDI or sheet music). However, audio-only approaches often struggle with polyphonic instruments and background noise. In contrast, visual information (e.g. a video of an instrument being played) does not have such ambiguities. In this work, we address the problem of transcribing piano music from visual data alone. We propose an end-to-end deep learning framework that learns to automatically predict note onset events given a video of a person playing the piano. From this, we are able to transcribe the played music in the form of MIDI data. We find that our approach is surprisingly effective in a variety of complex situations, particularly those in which music transcription from audio alone is impossible. We also show that combining audio and video data can improve the transcription obtained from each modality alone.

Correctly matching feature points in a pair of images is an important preprocessing step for many computer vision applications. In this paper we propose an efficient method for estimating the number of correct matches without explicitly computing them. To this end, we propose to analyze the set of matches using the spatial order of the features, as projected to the x-axis of the image. The set of features in each image is thus represented by a sequence, and analyzed using the Kendall and Spearman Footrule distance metrics between permutations. This result is interesting in its own right. Moreover, we demonstrate three useful applications of our method: (i) a new halting condition for RANSAC based epipolar geometry estimation methods, (ii) discarding spatially unrelated image pairs in the Structure-from-Motion pipeline, and (iii) computing the probability that a given match is correct based on the rank of the features within the sequences. Our experiments on a large number of synthetic and real data demonstrate the effectiveness of our method. For example, the running time of the image matching stage in the Structure-from-Motion pipeline may be reduced by about 90 percent while preserving about 85 percent of the image pairs with spatial overlap.

Dynamic events are often photographed by a number of people from different viewpoints at different times, resulting in an unconstrained set of images. Finding the corresponding moving features in each of the images allows us to extract information about objects of interest in the scene. Computing correspondence of moving features in such a set of images is considerably more challenging than computing correspondence in video due to possible significant differences in viewpoints and inconsistent timing between image captures. The prediction methods used in video for improving robustness and efficiency are not applicable to a set of still images. In this paper we propose a novel method to predict locations of an approximately linear moving feature point, given a small subset of correspondences and the temporal order of image captures. Our method extends the use of epipolar geometry to divide images into valid and invalid regions, termed Temporal Epipolar Regions (TERs). We formally prove that the location of a feature in a new image is restricted to valid TERs. We demonstrate the effectiveness of our method in reducing the search space for correspondence on both synthetic and challenging real world data, and show the improved matching.

Correctly matching feature points in a pair of images is an important preprocessing step for many computer vision applications. In this paper we propose an efficient method for estimating the number of correct matches without explicitly computing them. In addition, our method estimates the region of overlap between the images. To this end, we propose to analyze the set of matches using the spatial order of the features, as projected to the x-axis of the image. The set of features in each image is thus represented by a sequence. This reduces the analysis of the matching problem to the analysis of the permutation between the sequences. Using the Kendall distance metric between permutations and natural assumptions on the distribution of the correct and incorrect matches, we show how to estimate the abovementioned values. We demonstrate the usefulness of our method in two applications: (i) a new halting condition for RANSAC based epipolar geometry estimation methods that considerably reduce the running time, and (ii) discarding spatially unrelated image pairs in the Structure-from-Motion pipeline. Furthermore, our analysis can be used to compute the probability that a given match is correct based on the estimated number of correct matches and the rank of the features within the sequences. Our experiments on a large number of synthetic and real data demonstrate the effectiveness of our method. For example, the running time of the image matching stage in the Structure-from-Motion pipeline may be reduced by about 99% while preserving about 80% of the correctly matched feature points.

High-resolution images can be used to resolve matching ambiguities between trajectory fragments (tracklets), which is one of the main challenges in multiple target tracking. A PTZ camera, which can pan, tilt and zoom, is a powerful and efficient tool that offers both close-up views and wide area coverage on demand. The wide-area view makes it possible to track many targets while the close-up view allows individuals to be identified from high-resolution images of their faces. A central component of a PTZ tracking system is a scheduling algorithm that determines which target to zoom in on. In this paper we study this scheduling problem from a theoretical perspective, where the high resolution images are also used for tracklet matching. We propose a novel data structure, the Multi-Strand Tracking Graph (MSG), which represents the set of tracklets computed by a tracker and the possible associations between them. The MSG allows efficient scheduling as well as resolving - directly or by elimination - matching ambiguities between tracklets. The main feature of the MSG is the auxiliary data saved in each vertex, which allows efficient computation while avoiding time-consuming graph traversal. Synthetic data simulations are used to evaluate our scheduling algorithm and to demonstrate its superiority over a naïve one.

We propose a method to precisely segment books on bookshelves in images taken from general viewpoints. The proposed segmentation algorithm overcomes difficulties due to text and texture on book spines, various book orientations under perspective projection, and book proximity. A shape dependent active contour is used as a first step to establish a set of book spine candidates. A subset of these candidates are selected using spatial constraints on the assembly of spine candidates by formulating the selection problem as the maximal weighted independent set (MWIS) of a graph. The segmented book spines may be used by recognition systems (e.g., library automation), or rendered in computer graphics applications. We also propose a novel application that uses the segmented book spines to assist users in bookshelf reorganization or to modify the image to create a bookshelf with a tidier look. Our method was successfully tested on challenging sets of images.

Photo-sequencing is the problem of recovering the temporal order of a set of still images of a dynamic event, taken asynchronously by a set of uncalibrated cameras. Solving this problem is a first, crucial step for analyzing (or visualizing) the dynamic content of the scene captured by a large number of freely moving spectators. We propose a geometric based solution, followed by rank aggregation to the photo-sequencing problem. Our algorithm trades spatial certainty for temporal certainty. Whereas the previous solution proposed by [4] relies on two images taken from the same static camera to eliminate uncertainty in space, we drop the static-camera assumption and replace it with temporal information available from images taken from the same (moving) camera. Our method thus overcomes the limitation of the static-camera assumption, and scales much better with the duration of the event and the spread of cameras in space. We present successful results on challenging real data sets and large scale synthetic data (250 images).

Image retargeting algorithms attempt to adapt the image content to the screen without distorting the important objects in the scene. Existing methods address retargeting of a single image. In this paper, we propose a novel method for retargeting a pair of stereo images. Naively retargeting each image independently will distort the geometric structure and hence will impair the perception of the 3D structure of the scene. We show how to extend a single image seam carving to work on a pair of images. Our method minimizes the visual distortion in each of the images as well as the depth distortion. A key property of the proposed method is that it takes into account the visibility relations between pixels in the image pair (occluded and occluding pixels). As a result, our method guarantees, as we formally prove, that the retargeted pair is geometrically consistent with a feasible 3D scene, similar to the original one. Hence, the retargeted stereo pair can be viewed on a stereoscopic display or further processed by any computer vision algorithm. We demonstrate our method on a number of challenging indoor and outdoor stereo images.

Image retargeting algorithms attempt to adapt the image content to the screen without distorting the important objects in the scene. Existing methods address retargeting of a single image. In this paper we propose a novel method for retargeting a pair of stereo images. Naively retargeting each image independently will distort the geometric structure and make it impossible to perceive the 3D structure of the scene. We show how to extend a single image seam carving to work on a pair of images. Our method minimizes the visual distortion in each of the images as well as the depth distortion. A key property of the proposed method is that it takes into account the visibility relations between pixels in the image pair (occluded and occluding pixels). As a result, our method guarantees, as we formally prove, that the retargeted pair is geometrically consistent with a feasible 3D scene, similar to the original one. Hence, the retargeted stereo pair can be viewed on a stereoscopic display or processed by any computer vision algorithm. We demonstrate our method on a number of challenging indoor and outdoor stereo images.

We present a novel method for recovering the 3D structure and scene flow from calibrated multi-view sequences. We propose a 3D point cloud parametrization of the 3D structure and scene flow that allows us to directly estimate the desired unknowns. A unified global energy functional is proposed to incorporate the information from the available sequences and simultaneously recover both depth and scene flow. The functional enforces multi-view geometric consistency and imposes brightness constancy and piece-wise smoothness assumptions directly on the 3D unknowns. It inherently handles the challenges of discontinuities, occlusions, and large displacements. The main contribution of this work is the fusion of a 3D representation and an advanced variational framework that directly uses the available multi-view information. The minimization of the functional is successfully obtained despite the non-convex optimization problem. The proposed method was tested on real and synthetic data.

Many applications in computer vision require comparisons between two images of the same scene. Comparison applications usually assume that corresponding regions in the two images have similar colors. However, this assumption is not always true. One way to deal with this problem is to apply a color mapping to one of the images. In this paper we address the challenge of computing color mappings between pairs of images acquired under different acquisition conditions, and possibly by different cameras. For images taken from different viewpoints, our proposed method overcomes the lack of pixel correspondence. For images taken under different illumination, we show that no single color mapping exists, and we address and solve a new problem of computing a minimal set of piecewise color mappings. When both viewpoint and illumination vary, our method can only handle planar regions of the scene. In this case, the scene planar regions are simultaneously co-segmented in the two images, and piecewise color mappings for these regions are calculated. We demonstrate applications of the proposed method for each of these cases.

We present a new method for recovering the 3D shape of a featureless smooth surface from three or more calibrated images illuminated by different light sources (three of them are independent). This method is unique in its ability to handle images taken from unconstrained perspective viewpoints and unconstrained illumination directions. The correspondence between such images is hard to compute and no other known method can handle this problem locally from a small number of images. Our method combines geometric and photometric information in order to recover dense correspondence between the images and accurately computes the 3D shape. Only a single pass starting at one point and local computation are used. This is in contrast to methods that use the occluding contours recovered from many images to initialize and constrain an optimization process. The output of our method can be used to initialize such processes. In the special case of fixed viewpoint, the proposed method becomes a new perspective photometric stereo algorithm. Nevertheless, the introduction of the multiview setup, self-occlusions, and regions close to the occluding boundaries are better handled, and the method is more robust to noise than photometric stereo. Experimental results are presented for simulated and real images.

Tracking people in a dense crowd is a challenging problem for a single camera tracker due to occlusions and extensive motion that make human segmentation difficult. In this paper we suggest a method for simultaneously tracking all the people in a densely crowded scene using a set of cameras with overlapping fields of view. To overcome occlusions, the cameras are placed at a high elevation and only peoplepsilas heads are tracked. Head detection is still difficult since each foreground region may consist of multiple subjects. By combining data from several views, height information is extracted and used for head segmentation. The head tops, which are regarded as 2D patches at various heights, are detected by applying intensity correlation to aligned frames from the different cameras. The detected head tops are then tracked using common assumptions on motion direction and velocity. The method was tested on sequences in indoor and outdoor environments under challenging illumination conditions. It was successful in tracking up to 21 people walking in a small area (2.5 people per m2), in spite of severe and persistent occlusions.

One approach to recognizing objects seen from arbitrary viewpoint is by extracting invariant properties of the objects from single images. Such properties are found in images of 3D objects only when the objects are constrained to belong to certain classes (e.g., bilaterally symmetric objects). Existing studies that follow this approach propose how to compute invariant representations for a handful of classes of objects. A fundamental question regarding the invariance approach is whether it call be applied to a wide range of classes. To answer this question it is essential to study the set of classes for which invariance exists. This paper introduces a new method for determining the existence of invariance for classes of objects together with the set of images from which these invariance can be computed. We develop algebraic tests that, given a class of objects undergoing affine projection, determine whether the objects in the class can be identified from single images. In addition, these tests allow us to determine the sell of views of the objects which are degenerate. We apply these tests to several classes of objects and determine which of them is identifiable and which of their views are degenerate.

A face recognition system must recognize a face from a novel image despite the variations between images of the same face. A common approach to overcoming image variations because of changes in the illumination conditions is to use image representations that are relatively insensitive to these variations. Examples of such representations are edge maps, image intensity derivatives, and images convolved with 2D Gabor-like filters. Here we present an empirical study that evaluates the sensitivity of these representations to changes in illumination, as well as viewpoint and facial expression. Our findings indicated that none of the representations considered is sufficient by itself to overcome image variations because of a change in the direction of illumination. Similar results were obtained for changes due to viewpoint and expression. Image representations that emphasized the horizontal features were found to be less sensitive to changes in the direction of illumination. However, systems based only on such representations failed to recognize up to 20 percent of the faces in our database. Humans performed considerably better under the same conditions. We discuss possible reasons for this superiority and alternative methods for overcoming illumination effects in recognition.

We suggest an approach to describing and tracking the deformation of facial features. We concentrate on the mouth since its shape is important in detecting emotion. However, we believe that our system could be extended to deal with other facial features. In our system, the mouth is described by a valley contour which is based between the lips. This contour is shown to exist independently of illumination, viewpoint, identity, and expression. We present a real time mouth tracking system that follows this valley. It is shown to be robust to changes in identity, illumination and viewpoint. A simple classification algorithm was found to be sufficient to discriminate between 5 different mouth shapes, with a 100% recognition rate.>