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.

Modeling sensors: toward automatic generation of object recognition program

Abstract: One of the most important and systematic methods of building model-based vision systems is that of generating object recognition programs automatically from given geometric models. Automatic generation of object recognition programs requires several key components to be developed: object models to describe the geometric and photometric properties of the object to be recognized, sensor models to predict object appearances from the object model under a given sensor, strategy generation using the predicted appearances to produce a recognition strategy, and program generation converting the recognition strategy into an executable code. This paper concentrates on sensor modeling and its relationship to strategy generation, because we regard it as the bottleneck to automatic generation of object recognition programs. We consider two aspects of sensor characteristics: sensor detectability and sensor reliability. Sensor detectability specifies what kinds of featuers can be detected and under what conditions the features are detected; sensor reliability is a confidence level for the detected features. We define a configuration space to represent sensor characteristics. Then, we propose a representation method for sensor detectability and reliability in the configuration space. Finally, we investigate how to use the proposed sensor model in automatic generation of object recognition programs.

Incremental Acquisition of a Three-Dimensional Scene Model from Images

Abstract: We describe the current state of the 3-D Mosaic project, whose goal is to incrementally acquire a 3-D model of a complex urban scene from images. The notion of incremental acquisition arises from the observations that 1) single images contain only parfial information about a scene, 2) complex images are difficult to fully interpret, and 3) different features of a given scene tend to be easier to extract in different images because of differences in viewpoint and lighting conditions. In our approach, multiple images of the scene are sequentially analyzed so as to incrementaly construct the model. Each new image provides information which refines the model. We describe some experiments toward this end. Our method of extracting 3-D shape information from the images is stereo analysis. Because we are dealing with urban scenes, a junction-based matching technique proves very useful. This technique produces rather sparse wire-frame descriptions of the scene. A reasoning system that relies on task-specific knowledge generates an approximate model of the scene from the stereo output. Gray scale information is also acquired for the faces in the model. Finally, we describe an experiment in combining two views of the scene to obtain a rermed model.

Transgraph: interactive intensity-based 2D/3D registration of X-ray and CT data

Abstract: This paper presents work towards a system for intra-operative registration of 3D CT data to 2D X-ray radiographs. The registration procedure involves iterative comparison of Digitally Reconstructed Radiographs (DRRs) with X-ray images acquired during surgery. A new data structure called a Transgraph permits rapid generation of DRRS, and greatly speeds up the registration process. The underlying data structures are described, and the registration algorithm is evaluated for application to an existing image guided radiosurgery system.

A Hierarchical Markov Random Field Model for Figure-Ground Segregation

Abstract: To segregate overlapping objects into depth layers requires the integration of local occlusion cues distributed over the entire image into a global percept. We propose to model this process using hierarchical Markov random field (HMRF), and suggest a broader view that clique potentials in MRF models can be used to encode any local decision rules. A topology-dependent multiscale hierarchy is used to introduce long range interaction. The operations within each level are identical across the hierarchy. The clique parameters that encode the relative importance of these decision rules are estimated using an optimization technique called learning from rehearsals based on 2-object training samples. We find that this model generalizes successfully to 5-object test images, and that depth segregation can be completed within two traversals across the hierarchy. This computational framework therefore provides an interesting platform for us to investigate the interaction of local decision rules and global representations, as well as to reason about the rationales underlying some of recent psychological and neurophysiological findings related to figure-ground segregation.

A prediction and planning framework for road safety analysis, obstacle avoidance and driver information

Abstract: This paper presents a prediction and planning framework for analysing the safety and interaction of moving objects in complex road scenes. Rather than detecting specific, known, dangerous configurations, we simulate all the possible motion and interaction of objects. This simulation is used to detect dangerous situations, and to select the best path. The best path can be chosen according to a number of different criterion, such as: smoothest motion, largest avoiding distance, or quickest path. This framework can be applied, either as a driver warning system (open loop), or as an action recommendation system (human in the loop), or as an intelligent cruise control system (closed loop). This framework is evaluated using synthetic data, using simple and complex road scenes.

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.