Learning object segmentation from video data

This memo describes the initial results of a project to create a self-supervised algorithm for learning object segmentation from video data. Developmental psychology and computational experience have demonstrated that the motion segmentation of objects is a simpler, more primitive process than the detection of object boundaries by static image cues. Therefore, motion information provides a plausible supervision signal for learning the static boundary detection task and for evaluating performance on a test set. A video camera and previously developed background subtraction algorithms can automatically produce a large database of motion-segmented images for minimal cost. The purpose of this work is to use the information in such a database to learn how to detect the object boundaries in novel images using static information, such as color, texture, and shape. This work was funded in part by the Office of Naval Research contract #N00014-00-1-0298, in part by the Singapore-MIT Alliance agreement of 11/6/98, and in part by a National Science Foundation Graduate Student Fellowship.

Range Segmentation Using Visibility Constraints

Visibility constraints can aid the segmentation of foreground objects observed with multiple range images. In our approach, points are defined as foreground if they can be determined to occlude some {em empty space} in the scene. We present an efficient algorithm to estimate foreground points in each range view using explicit epipolar search. In cases where the background pattern is stationary, we show how visibility constraints from other views can generate virtual background values at points with no valid depth in the primary view. We demonstrate the performance of both algorithms for detecting people in indoor office environments.

Attentional Selection in Object Recognition

A key problem in object recognition is selection, namely, the problem of identifying regions in an image within which to start the recognition process, ideally by isolating regions that are likely to come from a single object. Such a selection mechanism has been found to be crucial in reducing the combinatorial search involved in the matching stage of object recognition. Even though selection is of help in recognition, it has largely remained unsolved because of the difficulty in isolating regions belonging to objects under complex imaging conditions involving occlusions, changing illumination, and object appearances. This thesis presents a novel approach to the selection problem by proposing a computational model of visual attentional selection as a paradigm for selection in recognition. In particular, it proposes two modes of attentional selection, namely, attracted and pay attention modes as being appropriate for data and model-driven selection in recognition. An implementation of this model has led to new ways of extracting color, texture and line group information in images, and their subsequent use in isolating areas of the scene likely to contain the model object. Among the specific results in this thesis are: a method of specifying color by perceptual color categories for fast color region segmentation and color-based localization of objects, and a result showing that the recognition of texture patterns on model objects is possible under changes in orientation and occlusions without detailed segmentation. The thesis also presents an evaluation of the proposed model by integrating with a 3D from 2D object recognition system and recording the improvement in performance. These results indicate that attentional selection can significantly overcome the computational bottleneck in object recognition, both due to a reduction in the number of features, and due to a reduction in the number of matches during recognition using the information derived during selection. Finally, these studies have revealed a surprising use of selection, namely, in the partial solution of the pose of a 3D object.

Feature Matching for Object Localization in the Presence of Uncertainty

We consider the problem of matching model and sensory data features in the presence of geometric uncertainty, for the purpose of object localization and identification. The problem is to construct sets of model feature and sensory data feature pairs that are geometrically consistent given that there is uncertainty in the geometry of the sensory data features. If there is no geometric uncertainty, polynomial-time algorithms are possible for feature matching, yet these approaches can fail when there is uncertainty in the geometry of data features. Existing matching and recognition techniques which account for the geometric uncertainty in features either cannot guarantee finding a correct solution, or can construct geometrically consistent sets of feature pairs yet have worst case exponential complexity in terms of the number of features. The major new contribution of this work is to demonstrate a polynomial-time algorithm for constructing sets of geometrically consistent feature pairs given uncertainty in the geometry of the data features. We show that under a certain model of geometric uncertainty the feature matching problem in the presence of uncertainty is of polynomial complexity. This has important theoretical implications by demonstrating an upper bound on the complexity of the matching problem, an by offering insight into the nature of the matching problem itself. These insights prove useful in the solution to the matching problem in higher dimensional cases as well, such as matching three-dimensional models to either two or three-dimensional sensory data. The approach is based on an analysis of the space of feasible transformation parameters. This paper outlines the mathematical basis for the method, and describes the implementation of an algorithm for the procedure. Experiments demonstrating the method are reported.

The Effect of Indexing on the Complexity of Object Recognition

Many current recognition systems use constrained search to locate objects in cluttered environments. Previous formal analysis has shown that the expected amount of search is quadratic in the number of model and data features, if all the data is known to come from a sinlge object, but is exponential when spurious data is included. If one can group the data into subsets likely to have come from a single object, then terminating the search once a "good enough" interpretation is found reduces the expected search to cubic. Without successful grouping, terminated search is still exponential. These results apply to finding instances of a known object in the data. In this paper, we turn to the problem of selecting models from a library, and examine the combinatorics of determining that a candidate object is not present in the data. We show that the expected search is again exponential, implying that naﶥ approaches to indexing are likely to carry an expensive overhead, since an exponential amount of work is needed to week out each of the incorrect models. The analytic results are shown to be in agreement with empirical data for cluttered object recognition.

Viewpoint-Specific Representations in Three-Dimensional Object Recognition

We report a series of psychophysical experiments that explore different aspects of the problem of object representation and recognition in human vision. Contrary to the paradigmatic view which holds that the representations are three-dimensional and object-centered, the results consistently support the notion of view-specific representations that include at most partial depth information. In simulated experiments that involved the same stimuli shown to the human subjects, computational models built around two-dimensional multiple-view representations replicated our main psychophysical results, including patterns of generalization errors and the time course of perceptual learning.

Some Extensions of the K-Means Algorithm for Image Segmentation and Pattern Classification

In this paper we present some extensions to the k-means algorithm for vector quantization that permit its efficient use in image segmentation and pattern classification tasks. It is shown that by introducing state variables that correspond to certain statistics of the dynamic behavior of the algorithm, it is possible to find the representative centers fo the lower dimensional maniforlds that define the boundaries between classes, for clouds of multi-dimensional, mult-class data; this permits one, for example, to find class boundaries directly from sparse data (e.g., in image segmentation tasks) or to efficiently place centers for pattern classification (e.g., with local Gaussian classifiers). The same state variables can be used to define algorithms for determining adaptively the optimal number of centers for clouds of data with space-varying density. Some examples of the applicatin of these extensions are also given.

Object Recognition By Alignment Using Invariant Projections of Planar Surfaces

In order to recognize an object in an image, we must determine the best transformation from object model to the image. In this paper, we show that for features from coplanar surfaces which undergo linear transformations in space, there exist projections invariant to the surface motions up to rotations in the image field. To use this property, we propose a new alignment approach to object recognition based on centroid alignment of corresponding feature groups. This method uses only a single pair of 2D model and data. Experimental results show the robustness of the proposed method against perturbations of feature positions.

View-Based Models of 3D Object Recognition and Class-Specific Invariances

This paper describes the main features of a view-based model of object recognition. The model tries to capture general properties to be expected in a biological architecture for object recognition. The basic module is a regularization network in which each of the hidden units is broadly tuned to a specific view of the object to be recognized.

Viewer-Centered Object Recognition in Monkeys

How does the brain recognize three-dimensional objects? We trained monkeys to recognize computer rendered objects presented from an arbitrarily chosen training view, and subsequently tested their ability to generalize recognition for other views. Our results provide additional evidence in favor of with a recognition model that accomplishes view-invariant performance by storing a limited number of object views or templates together with the capacity to interpolate between the templates (Poggio and Edelman, 1990).

Linear Object Classes and Image Synthesis from a Single Example Image

The need to generate new views of a 3D object from a single real image arises in several fields, including graphics and object recognition. While the traditional approach relies on the use of 3D models, we have recently introduced techniques that are applicable under restricted conditions but simpler. The approach exploits image transformations that are specific to the relevant object class and learnable from example views of other "prototypical" objects of the same class. In this paper, we introduce such a new technique by extending the notion of linear class first proposed by Poggio and Vetter. For linear object classes it is shown that linear transformations can be learned exactly from a basis set of 2D prototypical views. We demonstrate the approach on artificial objects and then show preliminary evidence that the technique can effectively "rotate" high- resolution face images from a single 2D view.

Spatial Reference Frames for Object Recognition: Tuning for Rotations in Depth

The inferior temporal cortex (IT) of monkeys is thought to play an essential role in visual object recognition. Inferotemporal neurons are known to respond to complex visual stimuli, including patterns like faces, hands, or other body parts. What is the role of such neurons in object recognition? The present study examines this question in combined psychophysical and electrophysiological experiments, in which monkeys learned to classify and recognize novel visual 3D objects. A population of neurons in IT were found to respond selectively to such objects that the monkeys had recently learned to recognize. A large majority of these cells discharged maximally for one view of the object, while their response fell off gradually as the object was rotated away from the neuron"s preferred view. Most neurons exhibited orientation-dependent responses also during view-plane rotations. Some neurons were found tuned around two views of the same object, while a very small number of cells responded in a view- invariant manner. For five different objects that were extensively used during the training of the animals, and for which behavioral performance became view-independent, multiple cells were found that were tuned around different views of the same object. No selective responses were ever encountered for views that the animal systematically failed to recognize. The results of our experiments suggest that neurons in this area can develop a complex receptive field organization as a consequence of extensive training in the discrimination and recognition of objects. Simple geometric features did not appear to account for the neurons" selective responses. These findings support the idea that a population of neurons -- each tuned to a different object aspect, and each showing a certain degree of invariance to image transformations -- may, as an assembly, encode complex 3D objects. In such a system, several neurons may be active for any given vantage point, with a single unit acting like a blurred template for a limited neighborhood of a single view.

Learning Object-Independent Modes of Variation with Feature Flow Fields

We present a unifying framework in which "object-independent" modes of variation are learned from continuous-time data such as video sequences. These modes of variation can be used as "generators" to produce a manifold of images of a new object from a single example of that object. We develop the framework in the context of a well-known example: analyzing the modes of spatial deformations of a scene under camera movement. Our method learns a close approximation to the standard affine deformations that are expected from the geometry of the situation, and does so in a completely unsupervised (i.e. ignorant of the geometry of the situation) fashion. We stress that it is learning a "parameterization", not just the parameter values, of the data. We then demonstrate how we have used the same framework to derive a novel data-driven model of joint color change in images due to common lighting variations. The model is superior to previous models of color change in describing non-linear color changes due to lighting.

Context-Based Vision System for Place and Object Recognition

While navigating in an environment, a vision system has to be able to recognize where it is and what the main objects in the scene are. In this paper we present a context-based vision system for place and object recognition. The goal is to identify familiar locations (e.g., office 610, conference room 941, Main Street), to categorize new environments (office, corridor, street) and to use that information to provide contextual priors for object recognition (e.g., table, chair, car, computer). We present a low-dimensional global image representation that provides relevant information for place recognition and categorization, and how such contextual information introduces strong priors that simplify object recognition. We have trained the system to recognize over 60 locations (indoors and outdoors) and to suggest the presence and locations of more than 20 different object types. The algorithm has been integrated into a mobile system that provides real-time feedback to the user.

Sharing visual features for multiclass and multiview object detection

We consider the problem of detecting a large number of different classes of objects in cluttered scenes. Traditional approaches require applying a battery of different classifiers to the image, at multiple locations and scales. This can be slow and can require a lot of training data, since each classifier requires the computation of many different image features. In particular, for independently trained detectors, the (run-time) computational complexity, and the (training-time) sample complexity, scales linearly with the number of classes to be detected. It seems unlikely that such an approach will scale up to allow recognition of hundreds or thousands of objects. We present a multi-class boosting procedure (joint boosting) that reduces the computational and sample complexity, by finding common features that can be shared across the classes (and/or views). The detectors for each class are trained jointly, rather than independently. For a given performance level, the total number of features required, and therefore the computational cost, is observed to scale approximately logarithmically with the number of classes. The features selected jointly are closer to edges and generic features typical of many natural structures instead of finding specific object parts. Those generic features generalize better and reduce considerably the computational cost of an algorithm for multi-class object detection.