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

: This report reviews progress at Carnegie Mellon from August 16, 1991 to August 15, 1992 on research sponsored by DARPA, DoD, monitored by the U.S. Army Topographic Engineering Center under contract DACA 76-89-C-0014, titled Perception for Outdoor Navigation. Research supported by this contract includes perception for road following, terrain mapping for off-road navigation, and systems software for building integrated mobile robots. We overview our efforts for the year, and list our publications and personnel, then provide further detail on several of our subprojects.

Third Annual Report for Perception for Outdoor Navigation

This paper examines the capturability of a guidance law that provides a desired angular acceleration of the heading angle to a pursuer. The guidance law has originally been proposed by Fajen and Warren[13], in order to explain the behavior of walking humans. It is shown that the relative rotational motion between a pursuer and an evader can be represented by an equation similar to the equation of motion for a pendulum with a disturbance. A set of sufficient conditions for successful pursuit are derived, and it is proved that the gain parameters of the guidance law can be chosen so that the conditions are satisfied. For sufficiently large gains, a pursuer controlled by the guidance law catches a maneuvering evader while achieving motion camouflage.

Capturability of a simple guidance law with angular acceleration input

Fat and muscle layers in the human soft-tissue causes degradation of resolution and distortions of shape in ultrasound images of organs. Novel ultrasonic applications such as registration between CT and ultrasound images, requiring comparable levels of accuracies between modalities could suffer from the distortions in ultrasound images. In this work, we describe a possible method to compensate or correct for these distortions in ultrasound images. This can be used as a pre-processing step before registration with other modalities. We derived a model of ultrasound field after reflection from an interface embedded in two layers. Experiments were performed using single-element transducers and custom made tissue mimicking phantoms of fat and muscle and a steel-block. We fit the model to the experimental data using Levenberg-Marquardt inversion.

/pdf/modeling-and-analysis-of-ultrasonic-echoes-reflected-from-a-by23qfy926.pdf

Modeling and Analysis of Ultrasonic Echoes Reflected from a Surface under Two Layers

From a real chair to a negative chair

Nighttime images of a scene from a surveillance camera have lower contrast and higher noise than their corresponding daytime images of the same scene due to low illumination. Denighting is an image enhancement method for improving nighttime images, so that they are closer to those that would have been taken during daytime. The method exploits the fact that background images of the same scene have been captured all day long with a much higher quality. We present several results of the enhancement of low quality nighttime images using denighting.

Takeo Kanade

Papers

Third Annual Report for Perception for Outdoor Navigation

Capturability of a simple guidance law with angular acceleration input

Modeling and Analysis of Ultrasonic Echoes Reflected from a Surface under Two Layers

From a real chair to a negative chair

Denight: nighttime image enhancement using daytime image