We seek to reconstruct sharp and noise-free high-dynamic range (HDR) video from a dual-exposure sensor that records different low-dynamic range (LDR) information in different pixel columns: Odd columns provide low-exposure, sharp, but noisy information; even columns complement this with less noisy, high-exposure, but motion-blurred data. Previous LDR work learns to deblur and denoise (DISTORTED->CLEAN) supervised by pairs of CLEAN and DISTORTED images. Regrettably, capturing DISTORTED sensor readings is time-consuming; as well, there is a lack of CLEAN HDR videos. We suggest a method to overcome those two limitations. First, we learn a different function instead: CLEAN->DISTORTED, which generates samples containing correlated pixel noise, and row and column noise, as well as motion blur from a low number of CLEAN sensor readings. Second, as there is not enough CLEAN HDR video available, we devise a method to learn from LDR video in-stead. Our approach compares favorably to several strong baselines, and can boost existing methods when they are re-trained on our data. Combined with spatial and temporal super-resolution, it enables applications such as re-lighting with low noise or blur.
Previous work has demonstrated learning isolated 3D objects (voxel grids, point clouds, meshes, etc.) from 2D-only self-supervision. We here set out to extend this to entire 3D scenes made out of multiple objects, including their location, orientation and type, and the scenes illumination. Once learned, we can map arbitrary 2D images to 3D scene structure. We analyze why analysis-by-synthesis-like losses for supervision of 3D scene structure using differentiable rendering is not practical, as it almost always gets stuck in local minima of visual ambiguities. This can be overcome by a novel form of training: we use an additional network to steer the optimization itself to explore the full gamut of possible solutions i.e. to be curious, and hence, to resolve those ambiguities and find workable minima. The resulting system converts 2D images of different virtual or real images into complete 3D scenes, learned only from 2D images of those scenes.
The motion of picking up and placing an object in 3D space is full of subtle detail. Typically these motions are formed from the same constraints, optimizing for swiftness, energy efficiency, as well as physiological limits. Yet, even for identical goals, the motion realized is always subject to natural variation. To capture these aspects computationally, we suggest a deep generative model for human reach-and-place action, conditioned on a start and end position.We have captured a dataset of 600 such human 3D actions, to sample the 2x3-D space of 3D source and targets. While temporal variation is often modeled with complex learning machinery like recurrent neural networks or networks with memory or attention, we here demonstrate a much simpler approach that is convolutional in time and makes use of(periodic) temporal encoding. Provided a latent code and conditioned on start and end position, the model generates a complete 3D character motion in linear time as a sequence of convolutions. Our evaluation includes several ablations, analysis of generative diversity and applications.
We suggest to represent an X-Field -a set of 2D images taken across different view, time or illumination conditions, i.e., video, light field, reflectance fields or combinations thereof-by learning a neural network (NN) to map their view, time or light coordinates to 2D images. Executing this NN at new coordinates results in joint view, time or light interpolation. The key idea to make this workable is a NN that already knows the "basic tricks" of graphics (lighting, 3D projection, occlusion) in a hard-coded and differentiable form. The NN represents the input to that rendering as an implicit map, that for any view, time, or light coordinate and for any pixel can quantify how it will move if view, time or light coordinates change (Jacobian of pixel position with respect to view, time, illumination, etc.). Our X-Field representation is trained for one scene within minutes, leading to a compact set of trainable parameters and hence real-time navigation in view, time and illumination.
Proteins perform a large variety of functions in living organisms, thus playing a key role in biology. As of now, available learning algorithms to process protein data do not consider several particularities of such data and/or do not scale well for large protein conformations. To fill this gap, we propose two new learning operations enabling deep 3D analysis of large-scale protein data. First, we introduce a novel convolution operator which considers both, the intrinsic (invariant under protein folding) as well as extrinsic (invariant under bonding) structure, by using $n$-D convolutions defined on both the Euclidean distance, as well as multiple geodesic distances between atoms in a multi-graph. Second, we enable a multi-scale protein analysis by introducing hierarchical pooling operators, exploiting the fact that proteins are a recombination of a finite set of amino acids, which can be pooled using shared pooling matrices. Lastly, we evaluate the accuracy of our algorithms on several large-scale data sets for common protein analysis tasks, where we outperform state-of-the-art methods.
Massive semantic labeling is readily available for 2D images, but much harder to achieve for 3D scenes. Objects in 3D repositories like ShapeNet are labeled, but regrettably only in isolation, so without context. 3D scenes can be acquired by range scanners on city-level scale, but much fewer with semantic labels. Addressing this disparity, we introduce a new optimization procedure, which allows training for 3D detection with raw 3D scans while using as little as 5% of the object labels and still achieve comparable performance. Our optimization uses two networks. A scene network maps an entire 3D scene to a set of 3D object centers. As we assume the scene not to be labeled by centers, no classic loss, such as chamfer can be used to train it. Instead, we use another network to emulate the loss. This loss network is trained on a small labeled subset and maps a non-centered 3D object in the presence of distractions to its own center. This function is very similar - and hence can be used instead of - the gradient the supervised loss would have. Our evaluation documents competitive fidelity at a much lower level of supervision, respectively higher quality at comparable supervision. Supplementary material can be found at: https://dgriffiths3.github.io.
We propose a generative model of 2D and 3D natural textures with diversity, visual fidelity and at high computational efficiency. This is enabled by a family of methods that extend ideas from classic stochastic procedural texturing (Perlin noise) to learned, deep, non-linearities. The key idea is a hard-coded, tunable and differentiable step that feeds multiple transformed random 2D or 3D fields into an MLP that can be sampled over infinite domains. Our model encodes all exemplars from a diverse set of textures without a need to be re-trained for each exemplar. Applications include texture interpolation, and learning 3D textures from 2D exemplars.
We suggest representing light field (LF) videos as "one-off" neural networks (NN), i.e., a learned mapping from view-plus-time coordinates to high-resolution color values, trained on sparse views. Initially, this sounds like a bad idea for three main reasons: First, a NN LF will likely have less quality than a same-sized pixel basis representation. Second, only few training data, e.g., 9 exemplars per frame are available for sparse LF videos. Third, there is no generalization across LFs, but across view and time instead. Consequently, a network needs to be trained for each LF video. Surprisingly, these problems can turn into substantial advantages: Other than the linear pixel basis, a NN has to come up with a compact, non-linear i.e., more intelligent, explanation of color, conditioned on the sparse view and time coordinates. As observed for many NN however, this representation now is interpolatable: if the image output for sparse view coordinates is plausible, it is for all intermediate, continuous coordinates as well. Our specific network architecture involves a differentiable occlusion-aware warping step, which leads to a compact set of trainable parameters and consequently fast learning and fast execution.
We show that denoising of 3D point clouds can be learned unsupervised, directly from noisy 3D point cloud data only. This is achieved by extending recent ideas from learning of unsupervised image denoisers to unstructured 3D point clouds. Unsupervised image denoisers operate under the assumption that a noisy pixel observation is a random realization of a distribution around a clean pixel value, which allows appropriate learning on this distribution to eventually converge to the correct value. Regrettably, this assumption is not valid for unstructured points: 3D point clouds are subject to total noise, i. e., deviations in all coordinates, with no reliable pixel grid. Thus, an observation can be the realization of an entire manifold of clean 3D points, which makes a na\"ive extension of unsupervised image denoisers to 3D point clouds impractical. Overcoming this, we introduce a spatial prior term, that steers converges to the unique closest out of the many possible modes on a manifold. Our results demonstrate unsupervised denoising performance similar to that of supervised learning with clean data when given enough training examples - whereby we do not need any pairs of noisy and clean training data.
We develop PlatonicGAN to discover 3D structure of an object class from an unstructured collection of 2D images. The key idea is to learn a deep neural network that generates 3D shapes that are never objectionable to a discriminator looking only at its 2D projections, i.e. renderings of the generated volumes. Using such a 2D instead of a 3D discriminator allows tapping into massive 2D image collections instead of relying on much smaller 3D data sets. To establish constraints between 2D image observation and their 3D interpretation we suggest a family of rendering layers that are effectively back-propagatable. This family includes visual hull, absorption-only (akin to x-ray), and emission-absorption (that can resolve occlusion if multiple 3D points project to the same 2D pixel). These layers are studied both on synthetic and real data in an application to reconstruct of 3D shape from 2D images.