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 a method for determining affine and metric calibration of a camera with unchanging internal parameters undergoing planar motion. It is shown that affine calibration is recovered uniquely, and metric calibration up to a two fold ambiguity.

/pdf/self-calibration-from-image-triplets-100dbtaowz.pdf

Self-Calibration from Image Triplets

We focus on multi-modal fusion for egocentric action recognition, and propose a novel architecture for multi-modal temporal-binding, i.e. the combination of modalities within a range of temporal offsets. We train the architecture with three modalities -- RGB, Flow and Audio -- and combine them with mid-level fusion alongside sparse temporal sampling of fused representations. In contrast with previous works, modalities are fused before temporal aggregation, with shared modality and fusion weights over time. Our proposed architecture is trained end-to-end, outperforming individual modalities as well as late-fusion of modalities. 
We demonstrate the importance of audio in egocentric vision, on per-class basis, for identifying actions as well as interacting objects. Our method achieves state of the art results on both the seen and unseen test sets of the largest egocentric dataset: EPIC-Kitchens, on all metrics using the public leaderboard.

EPIC-Fusion: Audio-Visual Temporal Binding for Egocentric Action Recognition

We describe a method for automatically obtaining object representations suitable for retrieval from generic video shots. The object representation consists of an association of frame regions. These regions provide exemplars of the object's possible visual appearances.

Two ideas are developed: (i) associating regions within a single shot to represent a deforming object; (ii) associating regions from the multiple visual aspects of a 3D object, thereby implicitly representing 3D structure. For the association we exploit temporal continuity (tracking) and wide baseline matching of affine covariant regions.

In the implementation there are three areas of novelty: First, we describe a method to repair short gaps in tracks. Second, we show how to join tracks across occlusions (where many tracks terminate simultaneously). Third, we develop an affine factorization method that copes with motion degeneracy.

We obtain tracks that last throughout the shot, without requiring a 3D reconstruction. The factorization method is used to associate tracks into object-level groups, with common motion. The outcome is that separate parts of an object that are not simultaneously visible (such as the front and back of a car, or the front and side of a face) are associated together. In turn this enables object-level matching and recognition throughout a video.

We illustrate the method on the feature film "Groundhog Day." Examples are given for the retrieval of deforming objects (heads, walking people) and rigid objects (vehicles, locations).

/pdf/object-level-grouping-for-video-shots-1vsfqifb5l.pdf

Object Level Grouping for Video Shots

The objective of this paper is self-supervised learning from video, in particular for representations for action recognition. We make the following contributions: (i) We propose a new architecture and learning framework Memory-augmented Dense Predictive Coding (MemDPC) for the task. It is trained with a predictive attention mechanism over the set of compressed memories, such that any future states can always be constructed by a convex combination of the condense representations, allowing to make multiple hypotheses efficiently. (ii) We investigate visual-only self-supervised video representation learning from RGB frames, or from unsupervised optical flow, or both. (iii) We thoroughly evaluate the quality of learnt representation on four different downstream tasks: action recognition, video retrieval, learning with scarce annotations, and unintentional action classification. In all cases, we demonstrate state-of-the-art or comparable performance over other approaches with orders of magnitude fewer training data.

Memory-augmented Dense Predictive Coding for Video Representation Learning

We describe a method for obtaining the principal objects, characters and scenes in a video by measuring the reoccurrence of spatial configurations of viewpoint invariant features. We investigate two aspects of the problem: the scale of the configurations, and the similarity requirements for clustering configurations. The problem is challenging firstly because an object can undergo substantial changes in imaged appearance throughout a video (due to viewpoint and illumination change, and partial occlusion), and secondly because configurations are detected imperfectly, so that inexact patterns must be matched. The novelty of the method is that viewpoint invariant features are used to form the configurations, and that efficient methods from the text analysis literature are employed to reduce the matching complexity. Examples of 'mined' objects are shown for a feature length film and a sitcom.

/pdf/video-data-mining-using-configurations-of-viewpoint-50yvhpnbxh.pdf

Andrew Zisserman

Papers

Self-Calibration from Image Triplets

EPIC-Fusion: Audio-Visual Temporal Binding for Egocentric Action Recognition

Object Level Grouping for Video Shots

Memory-augmented Dense Predictive Coding for Video Representation Learning

Video data mining using configurations of viewpoint invariant regions