Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

/pdf/deep-residual-learning-for-image-recognition-b75jg61qrq.pdf

Deep Residual Learning for Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Very Deep Convolutional Networks for Large-Scale Image Recognition

The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called “ImageNet”, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.

/pdf/imagenet-a-large-scale-hierarchical-image-database-1e8f663s1h.pdf

ImageNet: A large-scale hierarchical image database

There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net .

/pdf/u-net-convolutional-networks-for-biomedical-image-2xoyjzggxv.pdf

U-Net: Convolutional Networks for Biomedical Image Segmentation

We describe an approach to video object retrieval which enables all shots containing the object to be returned in a manner, and with a speed, similar to a Google search for text. The object is specified by a user outlining it in an image, and the object is then delineated in the retrieved shots.

/pdf/efficient-object-retrieval-from-videos-43ovw8osqi.pdf

Efficient object retrieval from videos

Outline This part is partly a recapitulation and partly new material. Chapter 17 is the recapitulation. We return to two- and three-view geometry but now within a more general framework which naturally extends to four- and n -views. The fundamental projective relations over multiple views arise from the intersection of lines (back-projected from points) and planes (back-projected from lines). These intersection properties are represented by the vanishing of determinants formed from the camera matrices of the views. The fundamental matrix, the trifocal tensor, and a new tensor for four views – the quadrifocal tensor – arise naturally from these determinants as the multiple view tensors for two, three, and four views respectively. The tensors are what remains when the 3D structure and non-essential part of the camera matrices are eliminated. The tensors stop at four views. These tensors are unique for each set of views, and generate relationships which are multi-linear in the coordinates of the image measurements. The tensors can be computed from sets of image correspondences, and subsequently a camera matrix for each view can be computed from the tensor. Finally, the 3D structure can be computed from the retrieved cameras and image correspondences. Chapter 18 covers the computation of a reconstruction from multiple views. In particular the important factorization algorithm is given for reconstruction from affine views. It is important because the algorithm is optimal, but is also non-iterative.

Multiple View Geometry in Computer Vision: N-View Geometry

In multiple-image super-resolution, a high-resolution image is estimated from a number of lower-resolution images. This usually involves computing the parameters of a generative imaging model (such as geometric and photometric registration, and blur) and obtaining a MAP estimate by minimizing a cost function including an appropriate prior. Two alternative approaches are examined. First, both registrations and the super-resolution image are found simultaneously using a joint MAP optimization. Second, we perform Bayesian integration over the unknown image registration parameters, deriving a cost function whose only variables of interest are the pixel values of the super-resolution image. We also introduce a scheme to learn the parameters of the image prior as part of the super-resolution algorithm. We show examples on a number of real sequences including multiple stills, digital video, and DVDs of movies.

/pdf/overcoming-registration-uncertainty-in-image-super-3tu08nd74k.pdf

Overcoming registration uncertainty in image super-resolution: maximize or marginalize?

Introduction—towards a new framework for vision

The objective of this paper is to rectify any monocular image by computing a homography matrix that transforms it to a bird's eye (overhead) view. 
We make the following contributions: (i) we show that the homography matrix can be parameterised with only four parameters that specify the horizon line and the vertical vanishing point, or only two if the field of view or focal length is known; (ii) We introduce a novel representation for the geometry of a line or point (which can be at infinity) that is suitable for regression with a convolutional neural network (CNN); (iii) We introduce a large synthetic image dataset with ground truth for the orthogonal vanishing points, that can be used for training a CNN to predict these geometric entities; and finally (iv) We achieve state-of-the-art results on horizon detection, with 74.52% AUC on the Horizon Lines in the Wild dataset. Our method is fast and robust, and can be used to remove perspective distortion from videos in real time.

Andrew Zisserman

Papers

Efficient object retrieval from videos

Multiple View Geometry in Computer Vision: N-View Geometry

Overcoming registration uncertainty in image super-resolution: maximize or marginalize?

Introduction—towards a new framework for vision

A Geometric Approach to Obtain a Bird's Eye View from an Image