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

Current methods for image segmentation are confused by artifacts such as highlights, because they are not based on any physical model of these phenomena. In this paper, we present an approach to color image understanding that accounts for color variations due to highlights and shading. Based on the physics of reflection by dielectric materials, such as plastic, we show that the color of every pixel from an object can be described as a linear combination of the object color and the highlight color. According to this model, all color pixels from one object form a planar cluster in the color space whose shape is determined by the object and highlight colors and by the object shape and illumination geometry. We present a method which exploits the color difference between object color and highlight color, as exhibited in the cluster shape, to separate the color of every pixel into a matte component and a highlight component. This generates two intrinsic images, one showing the scene without highlights, and the other one showing only the highlights. The intrinsic images may be a useful tool for a variety of algorithms in computer vision that cannot detect or analyze highlights, such as stereo vision, motion analysis, shape from shading, and shape from highlights. We have applied this method to real images in a laboratory environment, and we show these results and discuss some of the pragmatic issues endemic to precision color imaging.

http://far.in.tum.de/pub/klinker1987iccv/klinker1987iccv.pdf

Using a color reflection model to separate highlights from object color

For humanoid robots to fully realize their biped potential in a three-dimensional world and step over, around or onto obstacles such as stairs, appropriate and efficient approaches to execution, planning and perception are required. To this end, we have accelerated a robust model-based three-dimensional tracking system by programmable graphics hardware to operate online at frame-rate during locomotion of a humanoid robot. The tracker recovers the full 6 degree-of- freedom pose of viewable objects relative to the robot. Leveraging the computational resources of the GPU for perception has enabled us to increase our tracker's robustness to the significant camera displacement and camera shake typically encountered during humanoid navigation. We have combined our approach with a footstep planner and a controller capable of adaptively adjusting the height of swing leg trajectories. The resulting integrated perception-planning-action system has allowed an HRP-2 humanoid robot to successfully and rapidly localize, approach and climb stairs, as well as to avoid obstacles during walking.

/pdf/gpu-accelerated-real-time-3d-tracking-for-humanoid-4symijqp3i.pdf

GPU-accelerated real-time 3D tracking for humanoid locomotion and stair climbing

Surgical simulation is a promising technology for training medical students and planning procedures One major requirement for these simulation systems is a method to generate realistic cuts through soft tissue models This paper describes methods for cutting through tetrahedral models of soft tissue The cutting surface follows the free form path of the user’s motion, and generates a minimal set of new elements to replace intersected tetrahedra Intersected elements are progressively cut to minimize the lag between the user’s motion and model modification A linear finite element model is used to model deformation of the soft tissue These cutting techniques coupled with a physically based deformation model increases the accuracy and applicability of a surgical simulation system

/pdf/modifying-soft-tissue-models-progressive-cutting-with-3916qmkvj8.pdf

Modifying Soft Tissue Models: Progressive Cutting with Minimal New Element Creation

The motion of a non-rigid scene over time imposes more constraints on its structure than those derived from images at a single time instant alone. An algorithm is presented for simultaneously recovering dense scene shape and scene flow (i.e. the instantaneous 3D motion at every point in the scene). The algorithm operates by carving away hexels, or points in the 6D space of all possible shapes and flows that are inconsistent with the images captures at either time instant, or across time. The recovered shape is demonstrated to be more accurate than that recovered using images at a single time instant. Applications of the combined scene shape and flow include motion capture for animation, retiming of videos, and non-rigid motion analysis.

/pdf/shape-and-motion-carving-in-6d-trae1vk308.pdf

Takeo Kanade

Papers

Using a color reflection model to separate highlights from object color

GPU-accelerated real-time 3D tracking for humanoid locomotion and stair climbing

Modifying Soft Tissue Models: Progressive Cutting with Minimal New Element Creation

Shape and motion carving in 6D

Three-Dimensional Machine Vision