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.

Gradient-based learning applied to document recognition

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.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

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 …

I and i

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.

/pdf/histograms-of-oriented-gradients-for-human-detection-3nteinziuq.pdf

Histograms of oriented gradients for human detection

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.

/pdf/rich-feature-hierarchies-for-accurate-object-detection-and-22fvjlggyg.pdf

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

We present a neural network-based face detection system. A retinally connected neural network examines small windows of an image, and decides whether each window contains a face. The system arbitrates between multiple networks to improve performance over a single network. We use a bootstrap algorithm for training, which adds false detections into the training set as training progresses. This eliminates the difficult task of manually selecting non-face training examples, which must be chosen to span the entire space of non-face images. Comparisons with another state-of-the-art face detection system are presented; our system has better performance in terms of detection and false-positive rates.

/pdf/human-face-detection-in-visual-scenes-266687tg6e.pdf

Human Face Detection in Visual Scenes

A distributed architecture articulated around the CODGER (communication database with geometric reasoning) knowledge database is described for a mobile robot system that includes both perception and navigation tools. Results are described for vision and navigation tests using a mobile testbed that integrates perception and navigation capabilities that are based on two types of vision algorithms: color vision for road following, and 3-D vision for obstacle detection and avoidance. The perception modules are integrated into a system that allows the vehicle to drive continuously in an actual outdoor environment. The resulting system is able to navigate continuously on roads while avoiding obstacles. >

Vision and Navigation for the Carnegie-Mellon Navlab

Carnegie Mellon's Informedia Digital Video Library project will establish a large, on-line digital video library featuring full-content and knowledge-based search and retrieval. Intelligent, automatic mechanisms will be developed to populate the library. Search and retrieval from digital video, audio, and text libraries will take place via desktop computer over local, metropolitan, and wide-area networks. The project's approach applies several techniques for content-based searching and video-sequence retrieval. Content is conveyed in both the narrative (speech and language) and the image. Only by the collaborative interaction of image, speech, and natural language understanding technology is it possible to successfully populate, segment, index, and search diverse video collections with satisfactory recall and precision. This collaborative interaction approach uniquely compensates for problems of interpretation and search in error-ridden and ambiguous data sets. The authors have focused the work on two corpuses. One is science documentaries and lectures, the other is broadcast news content with partial closed-captions. Further work will continue to improve the accuracy and performance of the underlying processing as well as explore performance issues related to Web-based access and interoperability with other digital video resources.

Intelligent access to digital video: Informedia project

Correlation is a very effective way to align intensity images We extend the correlation technique to point set registration using a method we call kernel correlation Kernel correlation is an affinity measure, and it is also a function of the point set entropy We define the point set registration problem as finding the maximum kernel correlation configuration of the the two point sets to be registered The new registration method has intuitive interpretations, simple to implement algorithm and easy to prove convergence property Our method shows favorable performance when compared with the iterative closest point (ICP) and EM-ICP methods

/pdf/a-correlation-based-approach-to-robust-point-set-2x8zwhlhu8.pdf

A Correlation-Based Approach to Robust Point Set Registration

In this paper, we describe an algorithm for object recognition that explicitly models and estimated the posterior probability function, P(object/image). We have chosen a functional form of the posterior probability function that captures the joint statistics of local appearance and position on the object as well as the statistics of local appearance in the visual world at large. We use a discrete representation of local appearance consisting of approximately 10/sup 6/ patterns. We compute an estimate of P(object/image) in closed form by counting the frequency of occurrence of these patterns over various sets of training images. We have used this method for detecting human faces from frontal and profile views. The algorithm for frontal views has shown a detection rate of 93.0% with 88 false alarms on a set of 125 images containing 483 faces combining the MIT test set of Sung and Poggio with the CMU test sets of Rowley, Baluja, and Kanade. The algorithm for detection of profile views has also demonstrated promising results.

/pdf/probabilistic-modeling-of-local-appearance-and-spatial-5g4zy36dkz.pdf

Takeo Kanade

Papers

Human Face Detection in Visual Scenes

Vision and Navigation for the Carnegie-Mellon Navlab

Intelligent access to digital video: Informedia project

A Correlation-Based Approach to Robust Point Set Registration

Probabilistic modeling of local appearance and spatial relationships for object recognition