Takeo Kanade

Bio: Takeo Kanade is an academic researcher from Carnegie Mellon University. The author has contributed to research in topics: Motion estimation & Image processing. The author has an hindex of 147, co-authored 799 publications receiving 103237 citations. Previous affiliations of Takeo Kanade include National Institute of Advanced Industrial Science and Technology & Hitachi.

Cooperative Multi-Sensor Video Surveillance

Abstract: Carnegie Mellon University (CMU) and the David Sarno Research Center (Sarno ) have begun a joint, integrated feasibility demonstration in the area of Video Surveillance and Monitoring (VSAM). The objective is to develop a cooperative, multi-sensor video surveillance system that provides continuous coverage over large battle eld areas. Image Understanding (IU) technologies will be developed to: 1) coordinate multiple sensors to seamlessly track moving targets over an extended area, 2) actively control sensor and platform parameters to track multiple moving targets, 3) integrate multisensor output with collateral data to maintain an evolving, scene-level representation of all targets and platforms, and 4) monitor the scene for unusual \trigger" events and activities. These technologies will be integrated into an experimental testbed to support evaluation, data collection, and demonstration of other VSAM technologies developed within the DARPA IU community.

Autonomous scene description with range imagery

Abstract: This paper presents a program to produce object-centered 3-dimensional descriptions starting from point-wise 3D range data obtained by a light-stripe rangefinder. A careful geometrical analysis shows that contours which appear in light-stripe range images can be classified into eight types, each with different characteristics in occluding vs occluded and different camera/illuminator relationships. Starting with detecting these contours in the iconic range image, the descriptions are generated moving up the hierarchy of contour, surface, object, to scene. We use conical and cylindrical surfaces as primitives. In this process, we exploit the fact that coherent relatinships, such as symmetry, and being coaxial, which are present among lower-level elements in the hierarchy allow us to hypothesize upper-level elements. The resultant descriptions are used for matching and recognizing objects. The analysis program has been applied to complex scenes containing cups, pans, and toy shovels.

Geometric constraint analysis and synthesis: methods for improving shape-based registration accuracy

Abstract: Shape-based registration is a process for estimating the transformation between two shape representations of an object. It is used in many image-guided surgical systems to establish a transformation between pre- and intra-operative coordinate systems. This paper describes several tools which are useful for improving the accuracy resulting from shape-based registration: constraint analysis, constraint synthesis, and online accuracy estimation. Constraint analysis provides a scalar measure of sensitivity which is well correlated with registration accuracy. This measure can be used as a criterion function by constraint synthesis, an optimization process which generates configurations of registration data which maximize expected accuracy. Online accuracy estimation uses a conventional root-mean-squared error measure coupled with constraint analysis to estimate an upper bound on true registration error. This paper demonstrates that registration accuracy can be significantly improved via application of these methods.

A Comparative Study of Alternative FACS Coding Algorithms

Abstract: Two groups were contracted to experiment with coding of FACS (Ekman & Friesen, 1978) action units on a common database. One group is ours at CMU and the University of Pittsburgh, and the other is at UCSD. The database is from Frank and Ekman (1997) who video-recorded an interrogation in which subjects lied or told the truth about a mock crime. Subjects were ethnically diverse, action units occurred during speech, and out-of-plane motion and occlusion from head motion and glasses were common. The video data were originally collected to answer substantive questions in psychology, and represent a substantial challenge to automated AU recognition. This report describes the results of automated facial expression analysis by the CMU/Pittsburgh group. An interdisciplinary team of consultants, who have combined expertise in computer vision and in facial analysis, will compare the results of this report with those in a separate report submitted by UCSD group.

Object detection using 2D spatial ordering constraints

Abstract: Object detection is challenging partly due to the limited discriminative power of local feature descriptors. We amend this limitation by incorporating spatial constraints among neighboring features. We propose a two-step algorithm. First, a feature together with its spatial neighbors forms a flexible feature template. Two feature templates can be compared more informatively than two individual features without knowing the 3D object model. A large portion of false matches can be excluded after the first step. In a second global matching step, object detection is formulated as a graph-matching problem. A model graph is constructed by applying Delaunay triangulation on the surviving features. The best matching graph in an input image is computed by finding the maximum a posterior (MAP) estimate of a binary Markov random field with triangular maximal clique. The optimization is solved by the max-product algorithm (a.k.a. belief propagation). Experiments on both rigid and non-rigid objects demonstrate the generality and efficacy of the proposed methods.

Gradient-based learning applied to document recognition

Abstract: Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

I and i

Abstract: There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality. Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

Histograms of oriented gradients for human detection

Abstract: We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Abstract: Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3%. Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also present experiments that provide insight into what the network learns, revealing a rich hierarchy of image features. Source code for the complete system is available at http://www.cs.berkeley.edu/~rbg/rcnn.