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

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

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---8x deeper than VGG nets 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 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Hairpins are a ubiquitous secondary structure motif in RNA molecules. Despite their simple structure, there is some debate over whether they fold in a two-state or multi-state manner. We have studied the folding of a small tetraloop hairpin using a serial version of replica exchange molecular dynamics on a distributed computing environment. On the basis of these simulations, we have identified a number of intermediates that are consistent with experimental results. We also find that folding is not simply the reverse of high-temperature unfolding and suggest that this may be a general feature of biomolecular folding.

http://fodava.gatech.edu/files/reports/FODAVA-08-01.pdf

Structural insight into RNA hairpin folding intermediates.

Aiming at improving the classification accuracy with limited numbers of labeled pixels in polarimetric synthetic aperture radar (PolSAR) image classification task, this paper presents a graph-based semisupervised deep learning model for PolSAR image classification. It models the PolSAR image as an undirected graph, where the nodes correspond to the labeled and unlabeled pixels, and the weighted edges represent similarities between the pixels. Upon the graph, we design an energy function incorporating a semisupervision term, a convolutional neural network (CNN) term, and a pairwise smoothness term. The employed CNN extracts abstract and data-driven polarimetric features and outputs class label predictions to the graph model. The semisupervision term enforces the category label constraints on the human-labeled pixels. The pairwise smoothness term encourages class label smoothness and the alignment of class label boundaries with the image edges. Starting from an initialized class label map generated based on $K$ -Wishart distribution hypothesis or superpixel segmentation of PauliRGB images, we iteratively and alternately optimize the defined energy function until it converges. We conducted experiments on two real benchmark PolSAR images, and extensive experiments demonstrated that our approach achieved the state-of-the-art results for PolSAR image classification.

A Graph-Based Semisupervised Deep Learning Model for PolSAR Image Classification

Many applications such as topology repair, model editing, surface parameterization, and feature recognition benefit from computing loops on surfaces that wrap around their 'handles' and 'tunnels'. Computing such loops while optimizing their geometric lengths is difficult. On the other hand, computing such loops without considering geometry is easy but may not be very useful. In this paper we strike a balance by computing topologically correct loops that are also geometrically relevant. Our algorithm is a novel application of the concepts from topological persistence introduced recently in computational topology. The usability of the computed loops is demonstrated with some examples in feature identification and topology simplification.

/pdf/computing-geometry-aware-handle-and-tunnel-loops-in-3d-4g43twjsh1.pdf

Computing geometry-aware handle and tunnel loops in 3D models

Surface based 3D shape analysis plays a fundamental role in computer vision and medical imaging. This work proposes to use optimal mass transport map for shape matching and comparison, focusing on two important applications including surface registration and shape space. The computation of the optimal mass transport map is based on Monge-Brenier theory, in comparison to the conventional method based on Monge-Kantorovich theory, this method significantly improves the efficiency by reducing computational complexity from $O(n^2)$ to $O(n)$ . For surface registration problem, one commonly used approach is to use conformal map to convert the shapes into some canonical space. Although conformal mappings have small angle distortions, they may introduce large area distortions which are likely to cause numerical instability thus resulting failures of shape analysis. This work proposes to compose the conformal map with the optimal mass transport map to get the unique area-preserving map, which is intrinsic to the Riemannian metric, unique, and diffeomorphic. For shape space study, this work introduces a novel Riemannian framework, Conformal Wasserstein Shape Space , by combing conformal geometry and optimal mass transport theory. In our work, all metric surfaces with the disk topology are mapped to the unit planar disk by a conformal mapping, which pushes the area element on the surface to a probability measure on the disk. The optimal mass transport provides a map from the shape space of all topological disks with metrics to the Wasserstein space of the disk and the pullback Wasserstein metric equips the shape space with a Riemannian metric. We validate our work by numerous experiments and comparisons with prior approaches and the experimental results demonstrate the efficiency and efficacy of our proposed approach.

Optimal Mass Transport for Shape Matching and Comparison

Adversarial domain adaptation (DA) has been an effective approach for learning domain-invariant features by adversarial training. In this paper, we propose a novel adversarial DA approach completely defined in spherical feature space, in which we define spherical classifier for label prediction and spherical domain discriminator for discriminating domain labels. To utilize pseudo-label robustly, we develop a robust pseudo-label loss in the spherical feature space, which weights the importance of estimated labels of target data by posterior probability of correct labeling, modeled by Gaussian-uniform mixture model in spherical feature space. Extensive experiments show that our method achieves state-of-the-art results, and also confirm effectiveness of spherical classifier, spherical discriminator and spherical robust pseudo-label loss.

/pdf/spherical-space-domain-adaptation-with-robust-pseudo-label-41ulxg3uet.pdf

Jian Sun

Papers

Structural insight into RNA hairpin folding intermediates.

A Graph-Based Semisupervised Deep Learning Model for PolSAR Image Classification

Computing geometry-aware handle and tunnel loops in 3D models

Optimal Mass Transport for Shape Matching and Comparison

Spherical Space Domain Adaptation With Robust Pseudo-Label Loss