Josef Sivic

Lost in quantization: Improving particular object retrieval in large scale image databases

The state of the art in visual object retrieval from large databases is achieved by systems that are inspired by text retrieval. A key component of these approaches is that local regions of images are characterized using high-dimensional descriptors which are then mapped to ldquovisual wordsrdquo selected from a discrete vocabulary.This paper explores techniques to map each visual region to a weighted set of words, allowing the inclusion of features which were lost in the quantization stage of previous systems. The set of visual words is obtained by selecting words based on proximity in descriptor space. We describe how this representation may be incorporated into a standard tf-idf architecture, and how spatial verification is modified in the case of this soft-assignment. We evaluate our method on the standard Oxford Buildings dataset, and introduce a new dataset for evaluation. Our results exceed the current state of the art retrieval performance on these datasets, particularly on queries with poor initial recall where techniques like query expansion suffer. Overall we show that soft-assignment is always beneficial for retrieval with large vocabularies, at a cost of increased storage requirements for the index.

Learning and Transferring Mid-level Image Representations Using Convolutional Neural Networks

Convolutional neural networks (CNN) have recently shown outstanding image classification performance in the large- scale visual recognition challenge (ILSVRC2012). The success of CNNs is attributed to their ability to learn rich mid-level image representations as opposed to hand-designed low-level features used in other image classification methods. Learning CNNs, however, amounts to estimating millions of parameters and requires a very large number of annotated image samples. This property currently prevents application of CNNs to problems with limited training data. In this work we show how image representations learned with CNNs on large-scale annotated datasets can be efficiently transferred to other visual recognition tasks with limited amount of training data. We design a method to reuse layers trained on the ImageNet dataset to compute mid-level image representation for images in the PASCAL VOC dataset. We show that despite differences in image statistics and tasks in the two datasets, the transferred representation leads to significantly improved results for object and action classification, outperforming the current state of the art on Pascal VOC 2007 and 2012 datasets. We also show promising results for object and action localization.

Using Multiple Segmentations to Discover Objects and their Extent in Image Collections

Given a large dataset of images, we seek to automatically determine the visually similar object and scene classes together with their image segmentation. To achieve this we combine two ideas: (i) that a set of segmented objects can be partitioned into visual object classes using topic discovery models from statistical text analysis; and (ii) that visual object classes can be used to assess the accuracy of a segmentation. To tie these ideas together we compute multiple segmentations of each image and then: (i) learn the object classes; and (ii) choose the correct segmentations. We demonstrate that such an algorithm succeeds in automatically discovering many familiar objects in a variety of image datasets, including those from Caltech, MSRC and LabelMe.

Hello! My name is... Buffy - Automatic Naming of Characters in TV Video

We investigate the problem of automatically labelling appearances of characters in TV or film material. This is tremendously challenging due to the huge variation in imaged appearance of each character and the weakness and ambiguity of available annotation. However, we demonstrate that high precision can be achieved by combining multiple sources of information, both visual and textual. The principal novelties that we introduce are: (i) automatic generation of time stamped character annotation by aligning subtitles and transcripts; (ii) strengthening the supervisory information by identifying when characters are speaking; (iii) using complementary cues of face matching and clothing matching to propose common annotations for face tracks. Results are presented on episodes of the TV series “Buffy the Vampire Slayer”.

SIFT Flow: Dense Correspondence across Different Scenes

While image registration has been studied in different areas of computer vision, aligning images depicting different scenes remains a challenging problem, closer to recognition than to image matching. Analogous to optical flow, where an image is aligned to its temporally adjacent frame, we propose SIFT flow, a method to align an image to its neighbors in a large image collection consisting of a variety of scenes. For a query image, histogram intersection on a bag-of-visual-words representation is used to find the set of nearest neighbors in the database. The SIFT flow algorithm then consists of matching densely sampled SIFT features between the two images, while preserving spatial discontinuities. The use of SIFT features allows robust matching across different scene/object appearances and the discontinuity-preserving spatial model allows matching of objects located at different parts of the scene. Experiments show that the proposed approach is able to robustly align complicated scenes with large spatial distortions. We collect a large database of videos and apply the SIFT flow algorithm to two applications: (i) motion field prediction from a single static image and (ii) motion synthesis via transfer of moving objects.

Non-uniform Deblurring for Shaken Images

Photographs taken in low-light conditions are often blurry as a result of camera shake, i.e. a motion of the camera while its shutter is open. Most existing deblurring methods model the observed blurry image as the convolution of a sharp image with a uniform blur kernel. However, we show that blur from camera shake is in general mostly due to the 3D rotation of the camera, resulting in a blur that can be significantly non-uniform across the image. We propose a new parametrized geometric model of the blurring process in terms of the rotational motion of the camera during exposure. This model is able to capture non-uniform blur in an image due to camera shake using a single global descriptor, and can be substituted into existing deblurring algorithms with only small modifications. To demonstrate its effectiveness, we apply this model to two deblurring problems; first, the case where a single blurry image is available, for which we examine both an approximate marginalization approach and a maximum a posteriori approach, and second, the case where a sharp but noisy image of the scene is available in addition to the blurry image. We show that our approach makes it possible to model and remove a wider class of blurs than previous approaches, including uniform blur as a special case, and demonstrate its effectiveness with experiments on synthetic and real images.

What makes Paris look like Paris

Given a large repository of geotagged imagery, we seek to automatically find visual elements, e. g. windows, balconies, and street signs, that are most distinctive for a certain geo-spatial area, for example the city of Paris. This is a tremendously difficult task as the visual features distinguishing architectural elements of different places can be very subtle. In addition, we face a hard search problem: given all possible patches in all images, which of them are both frequently occurring and geographically informative? To address these issues, we propose to use a discriminative clustering approach able to take into account the weak geographic supervision. We show that geographically representative image elements can be discovered automatically from Google Street View imagery in a discriminative manner. We demonstrate that these elements are visually interpretable and perceptually geo-informative. The discovered visual elements can also support a variety of computational geography tasks, such as mapping architectural correspondences and influences within and across cities, finding representative elements at different geo-spatial scales, and geographically-informed image retrieval.

Is object localization for free? - Weakly-supervised learning with convolutional neural networks

Successful methods for visual object recognition typically rely on training datasets containing lots of richly annotated images. Detailed image annotation, e.g. by object bounding boxes, however, is both expensive and often subjective. We describe a weakly supervised convolutional neural network (CNN) for object classification that relies only on image-level labels, yet can learn from cluttered scenes containing multiple objects. We quantify its object classification and object location prediction performance on the Pascal VOC 2012 (20 object classes) and the much larger Microsoft COCO (80 object classes) datasets. We find that the network (i) outputs accurate image-level labels, (ii) predicts approximate locations (but not extents) of objects, and (iii) performs comparably to its fully-supervised counterparts using object bounding box annotation for training.

Unsupervised discovery of visual object class hierarchies

Objects in the world can be arranged into a hierarchy based on their semantic meaning (e.g. organism - animal - feline - cat). What about defining a hierarchy based on the visual appearance of objects? This paper investigates ways to automatically discover a hierarchical structure for the visual world from a collection of unlabeled images. Previous approaches for unsupervised object and scene discovery focused on partitioning the visual data into a set of non-overlapping classes of equal granularity. In this work, we propose to group visual objects using a multi-layer hierarchy tree that is based on common visual elements. This is achieved by adapting to the visual domain the generative hierarchical latent Dirichlet allocation (hLDA) model previously used for unsupervised discovery of topic hierarchies in text. Images are modeled using quantized local image regions as analogues to words in text. Employing the multiple segmentation framework of Russell et al. [22], we show that meaningful object hierarchies, together with object segmentations, can be automatically learned from unlabeled and unsegmented image collections without supervision. We demonstrate improved object classification and localization performance using hLDA over the previous non-hierarchical method on the MSRC dataset [33].

Person spotting: video shot retrieval for face sets

Matching people based on their imaged face is hard because of the well known problems of illumination, pose, size and expression variation. Indeed these variations can exceed those due to identity. Fortunately, videos of people have the happy benefit of containing multiple exemplars of each person in a form that can easily be associated automatically using straightforward visual tracking. We describe progress in harnessing these multiple exemplars in order to retrieve humans automatically in videos, given a query face in a shot. There are three areas of interest: (i) the matching of sets of exemplars provided by “tubes” of the spatial-temporal volume; (ii) the description of the face using a spatial orientation field; and, (iii) the structuring of the problem so that retrieval is immediate at run time. The result is a person retrieval system, able to retrieve a ranked list of shots containing a particular person in the manner of Google. The method has been implemented and tested on two feature length movies.