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

Today, visual recognition systems are still rarely employed in robotics applications. Perhaps one of the main reasons for this is the lack of demanding benchmarks that mimic such scenarios. In this paper, we take advantage of our autonomous driving platform to develop novel challenging benchmarks for the tasks of stereo, optical flow, visual odometry/SLAM and 3D object detection. Our recording platform is equipped with four high resolution video cameras, a Velodyne laser scanner and a state-of-the-art localization system. Our benchmarks comprise 389 stereo and optical flow image pairs, stereo visual odometry sequences of 39.2 km length, and more than 200k 3D object annotations captured in cluttered scenarios (up to 15 cars and 30 pedestrians are visible per image). Results from state-of-the-art algorithms reveal that methods ranking high on established datasets such as Middlebury perform below average when being moved outside the laboratory to the real world. Our goal is to reduce this bias by providing challenging benchmarks with novel difficulties to the computer vision community. Our benchmarks are available online at: www.cvlibs.net/datasets/kitti

/pdf/are-we-ready-for-autonomous-driving-the-kitti-vision-zygefho352.pdf

Are we ready for autonomous driving? The KITTI vision benchmark suite

Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, exceed the state-of-the-art in semantic segmentation. Our key insight is to build "fully convolutional" networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning. We define and detail the space of fully convolutional networks, explain their application to spatially dense prediction tasks, and draw connections to prior models. We adapt contemporary classification networks (AlexNet, the VGG net, and GoogLeNet) into fully convolutional networks and transfer their learned representations by fine-tuning to the segmentation task. We then define a novel architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations. Our fully convolutional network achieves state-of-the-art segmentation of PASCAL VOC (20% relative improvement to 62.2% mean IU on 2012), NYUDv2, and SIFT Flow, while inference takes one third of a second for a typical image.

Fully Convolutional Networks for Semantic Segmentation

Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption.

/pdf/pointnet-deep-learning-on-point-sets-for-3d-classification-1zge9sjodi.pdf

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation

A survey on deep learning in medical image analysis

In this work we consider the regularization of vectorial data such as color images. Based on the observation that edge alignment across image channels is a desirable prior for multichannel image restoration, we propose a novel scheme of minimizing the rank of the image Jacobian and extend this idea to second derivatives in the framework of total generalized variation. We compare the proposed convex and nonconvex relaxations of the rank function based on the Schatten-q norm to previous color image regularizers and show in our numerical experiments that they have several desirable properties. In particular, the nonconvex relaxations lead to better preservation of discontinuities. The efficient minimization of energies involving nonconvex and nonsmooth regularizers is still an important open question. We extend a recently proposed primal-dual splitting approach for nonconvex optimization and show that it can be effectively used to minimize such energies. Furthermore, we propose a novel algorithm for efficiently evaluating the proximal mapping of the l q norm appearing during optimization. We experimentally verify convergence of the proposed optimization method and show that it performs comparably to sequential convex programming.

/pdf/low-rank-priors-for-color-image-regularization-5fli3uo87c.pdf

Low Rank Priors for Color Image Regularization

The introduction of statistical shape knowledge into level set based segmentation methods was shown to improve the segmentation of familiar structures in the presence of noise, clutter or partial occlusions. While most work has been focused on shape priors which are constant in time, it is clear that when tracking deformable shapes certain silhouettes may become more or less likely over time. In fact, the deformations of familiar objects such as the silhouettes of a walking person are often characterized by pronounced temporal correlations. In this paper, we propose a nonlinear dynamical shape prior for level set based image segmentation. Specifically, we propose to approximate the temporal evolution of the eigenmodes of the level set function by means of a mixture of autoregressive models. We detail how such shape priors "with memory" can be integrated into a variational framework for level set segmentation. As an application, we experimentally validate that the nonlinear dynamical prior drastically improves the tracking of a person walking in different directions, despite large amounts of clutter and noise.

/pdf/nonlinear-dynamical-shape-priors-for-level-set-segmentation-481zpp9nsm.pdf

Nonlinear Dynamical Shape Priors for Level Set Segmentation

Traditional optical flow algorithms rely on consecutive short-exposed images. In this work, we make use of an additional long-exposed image for motion field estimation. Long-exposed images integrate motion information directly in the form of motion-blur. With this additional information, more robust and accurate motion fields can be estimated. In addition, the moment of occlusion can be determined. Considering the basic signal-theoretical problem in motion field estimation, we exploit the fact that long-exposed images integrate motion information to prevent temporal aliasing. A suitable image formation model relates the long-exposed image to preceding and succeeding short-exposed images in terms of dense 2D motion and per-pixel occlusion/disocclusion timings. Based on our image formation model, we describe a practical variational algorithm to estimate the motion field not only for visible image regions but also for regions getting occluded. Results for synthetic as well as real-world scenes demonstrate the validity of the approach.

Motion Field Estimation from Alternate Exposure Images

We propose a probabilistic formulation of 3D segmentation given a series of images from calibrated cameras. Instead of segmenting each image separately in order to build a 3D surface consistent with these segmentations, we compute the most probable surface that gives rise to the images. Additionally, our method can reconstruct the mean intensity and variance of the extracted object and background. Although it is designed for scenes, where the objects can be distinguished visually from the background (i.e. images of piecewise homogeneous regions), the proposed algorithm can also cope with noisy data. We carry out the numerical implementation in the level set framework. Our experiments on synthetic data sets reveal favorable results compared to state-of-the-art methods, in particular in terms of robustness to noise and initialization.

/pdf/robust-variational-segmentation-of-3d-objects-from-multiple-3nwp8zw5b2.pdf

Robust variational segmentation of 3d objects from multiple views

We present a variational approach to jointly estimate a displacement map and a superresolution texture for a 3D model from multiple calibrated views. The superresolution image formation model leads to an energy functional defined in terms of an integral over the object surface. This functional can be minimized by alternately solving a deblurring PDE and a total variation minimization on the surface, leading to increasingly accurate estimates of photometry and geometry, respectively. The resulting equations can be discretized and solved on texture space with the help of a conformal atlas. The superresolution approach to texture reconstruction allows to obtain fine details in the texture map which surpass individual input image resolution.

https://www.tpbin.com/Uploads/Subjects/7639f3e9-aaaf-4b46-99c0-7223d58dea57.pdf

Daniel Cremers

Papers

Low Rank Priors for Color Image Regularization

Nonlinear Dynamical Shape Priors for Level Set Segmentation

Motion Field Estimation from Alternate Exposure Images

Robust variational segmentation of 3d objects from multiple views

A Superresolution Framework for High-Accuracy Multiview Reconstruction