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

The last decade has seen automatic face recognition evolve from small-scale research systems to a wide range of commercial products. Driven by the FERET face database and evaluation protocol, the currently best commercial systems achieve verification accuracies comparable to those of fingerprint recognizers. In these experiments, only frontal face images taken under controlled lighting conditions were used. As the use of face recognition systems expands toward less restricted environments, the development of algorithms for view and illumination invariant face recognition becomes important. However, the performance of current algorithms degrades significantly when tested across pose and illumination, as documented in a number of evaluations. In this chapter, we review previously proposed algorithms for pose and illumination invariant face recognition. We then describe in detail two successful appearance-based algorithms for face recognition across pose, eigen light-fields, and Bayesian face subregions. We furthermore show how both of these algorithms can be extended toward face recognition across pose and illumination.

/pdf/face-recognition-across-pose-and-illumination-wcagzyew7d.pdf

Face Recognition Across Pose and Illumination

Visual hull (VH) construction from silhouette images is a popular method of shape estimation. The method, also known as shape-from-silhouette (SFS), is used in many applications such as non-invasive 3D model acquisition, obstacle avoidance, and more recently human motion tracking and analysis. One of the limitations of SFS, however, is that the approximated shape can be very coarse when there are only a few cameras. In this paper, we propose an algorithm to improve the shape approximation by combining multiple silhouette images captured across time. The improvement is achieved by first estimating the rigid motion between the visual hulls formed at different time instants (visual hull alignment) and then combining them (visual hull refinement) to get a tighter bound on the object's shape. Our algorithm first constructs a representation of the VHs called the bounding edge representation. Utilizing a fundamental property of visual hulls, which states that each bounding edge must touch the object at at least one point, we use multi-view stereo to extract points called colored surface points (CSP) on the surface of the object. These CSPs are then used in a 3D image alignment algorithm to find the 6 DOF rigid motion between two visual hulls. Once the rigid motion across time is known, all of the silhouette images are treated as being captured at the same time instant and the shape of the object is refined. We validate our algorithm on both synthetic and real data and compare it with space carving.

/pdf/visual-hull-alignment-and-refinement-across-time-a-3d-2qt80t90i3.pdf

Visual hull alignment and refinement across time: a 3D reconstruction algorithm combining shape-from-silhouette with stereo

This paper proposes a novel approach to extract meaningful content information from video by collaborative integration of image understanding and natural language processing. As an actual example, we developed a system that associates faces and names in videos, called Name-It, which is given news videos as a knowledge source, then automatically extracts face and name association as content information. The system can infer the name of a given unknown face image, or guess faces which are likely to have the name given to the system. This paper explains the method with several successful matching results which reveal effectiveness in integrating heterogeneous techniques as well as the importance of real content information extraction from video, especially face-name association.

Name-It: association of face and name in video

We consider the problem of designing a scene-specific pedestrian detector in a scenario where we have zero instances of real pedestrian data (i.e., no labeled real data or unsupervised real data). This scenario may arise when a new surveillance system is installed in a novel location and a scene-specific pedestrian detector must be trained prior to any observations of pedestrians. The key idea of our approach is to infer the potential appearance of pedestrians using geometric scene data and a customizable database of virtual simulations of pedestrian motion. We propose an efficient discriminative learning method that generates a spatially-varying pedestrian appearance model that takes into the account the perspective geometry of the scene. As a result, our method is able to learn a unique pedestrian classifier customized for every possible location in the scene. Our experimental results show that our proposed approach outperforms classical pedestrian detection models and hybrid synthetic-real models. Our results also yield a surprising result, that our method using purely synthetic data is able to outperform models trained on real scene-specific data when data is limited.

/pdf/learning-scene-specific-pedestrian-detectors-without-real-4irrfv4mbx.pdf

Learning scene-specific pedestrian detectors without real data

In this paper, we present a robust face alignment system that is capable of dealing with exaggerating expressions, large occlusions, and a wide variety of image noises. The robustness comes from our shape regularization model, which incorporates constrained nonlinear shape prior, geometric transformation, and likelihood of multiple candidate landmarks in a three-layered generative model. The inference algorithm iteratively examines the best candidate positions and updates face shape and pose. This model can effectively recover sufficient shape details from very noisy observations. We demonstrate the performance of this approach on two public domain databases and a large collection of real-world face photographs.

/pdf/a-generative-shape-regularization-model-for-robust-face-3i81rtbzbc.pdf

Takeo Kanade

Papers

Face Recognition Across Pose and Illumination

Visual hull alignment and refinement across time: a 3D reconstruction algorithm combining shape-from-silhouette with stereo

Name-It: association of face and name in video

Learning scene-specific pedestrian detectors without real data

A Generative Shape Regularization Model for Robust Face Alignment