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

Image segmentation is essential for many automated microscopy image analysis systems. Rather than treating microscopy images as general natural images and rushing into the image processing warehouse for solutions, we propose to study a microscope's optical properties to model its image formation process first using phase contrast microscopy as an exemplar. It turns out that the phase contrast imaging system can be relatively well explained by a linear imaging model. Using this model, we formulate a quadratic optimization function with sparseness and smoothness regularizations to restore the "authentic" phase contrast images that directly correspond to specimen's optical path length without phase contrast artifacts such as halo and shade-off. With artifacts removed, high quality segmentation can be achieved by simply thresholding the restored images. The imaging model and restoration method are quantitatively evaluated on two sequences with thousands of cells captured over several days.

/pdf/understanding-the-optics-to-aid-microscopy-image-5f1s54wc0a.pdf

Understanding the optics to aid microscopy image segmentation

Clinical translation of stem cell research promises to revolutionize medicine. Challenges remain toward belter understanding of stem cell biology and cost-effective strategies for stem cell manufacturing. These challenges call for novel engineering toolsets to study stem cell behaviors and the associated sternness. Towards this goal, we are developing a computer vision based system to automatically and reliably follow the behaviors of individual stem cells in expanding populations. This paper reports on significant progress in our development. In particular, we present a machine-learning approach for detecting spatiotemporal mitosis events without image segmentation. This approach not only improves tracking performance, but can also independently quantify mitoses and cellular divisions. We also employ bilateral filtering to improve cell detection performance. We demonstrate the effectiveness of this system on tracking C2C12 mouse myoblast stem cells.

/pdf/computer-vision-tracking-of-stemness-4p6czy7tvg.pdf

Computer vision tracking of stemness

A new visual medium, Virtualized Reality, immerses viewers in a virtual reconstruction of real-world events. The Virtualized Reality world model consists of real images and depth information computed from these images. Stereoscopic reconstructions provide a sense of complete immersion, and users can select their own viewpoints at view time, independent of the actual camera positions used to capture the event.

Virtualized reality : Constructing virtual worlds from real scenes : Immersive telepresence

We have designed a physical awareness system called CAMEO, the camera assisted meeting event observer, which consists of a multi-camera omnidirectional vision system designed to be used in meeting environments. CAMEO is designed to monitor the activities of people in meetings so that it can generate a semantically-indexed summary of what occurred in the meeting. In this paper, we describe CAMEO's fast people detection and tracking module. This module makes use of a combination of frame differencing, face detection, and adaptive color blob tracking based on mean shift analysis to detect and track people in the panoramic image. We describe this algorithm and present experimental results from captured meeting logs.

/pdf/people-detection-and-tracking-in-high-resolution-panoramic-1n9gqq1l0f.pdf

Takeo Kanade

Papers

Understanding the optics to aid microscopy image segmentation

Computer vision tracking of stemness

Virtualized reality : Constructing virtual worlds from real scenes : Immersive telepresence

People detection and tracking in high resolution panoramic video mosaic

A Method of Time-Coded Parallel Planes of Light for Depth Measurement