Motion control is crucial for generating expressive and compelling video content; however, most existing video generation models rely mainly on text prompts for control, which struggle to capture the nuances of dynamic actions and temporal compositions. To this end, we train a video generation model conditioned on spatiotemporally sparse or dense motion trajectories. In contrast to prior motion conditioning work, this flexible representation can encode any number of trajectories, object-specific or global scene motion, and temporally sparse motion; due to its flexibility we refer to this conditioning as motion prompts. While users may directly specify sparse trajectories, we also show how to translate high-level user requests into detailed, semi-dense motion prompts, a process we term motion prompt expansion. We demonstrate the versatility of our approach through various applications, including camera and object motion control, "interacting" with an image, motion transfer, and image editing. Our results showcase emergent behaviors, such as realistic physics, suggesting the potential of motion prompts for probing video models and interacting with future generative world models. Finally, we evaluate quantitatively, conduct a human study, and demonstrate strong performance.

Phase unwrapping is a crucial step in deriving deformation from interferometric synthetic aperture radar data, yet challenges such as discontinuous noise in low-coherence areas and random sensor noise can significantly compromise reconstruction accuracy and efficiency. Existing methods often struggle under noisy conditions and exhibit slower processing times. To overcome these limitations, this paper introduces Phaseformer, a transformer-based model designed to predict the wrap count at each pixel directly from wrapped phase maps. Due to the limited size of the real-world dataset, a simulated dataset with diverse noise levels was developed to train the model. Phaseformer achieved the lowest RMSE of 1.88 rad, a success ratio of 95.6%, and a runtime of 0.01 s per image, outperforming both state-of-the-art methods and traditional methods. For real-world scenarios, unwrapped phase results were integrated into a post-processing workflow using MintPy to derive time-series deformation. Interferograms from Hawaii and Mexico City, generated using Sentinel-1 data and validated with GPS ground truth, revealed consistent deformation rates with minimal loop closure errors. These results highlight Phaseformer’s potential for accurate and efficient phase unwrapping in complex, real-world applications.

Video editing is a creative and complex endeavor and we believe that there is potential for reimagining a new video editing interface to better support the creative and exploratory nature of video editing. We take inspiration from latent space exploration tools that help users find patterns and connections within complex datasets. We present VideoMap, a proof-of-concept video editing interface that operates on video frames projected onto a latent space. We support intuitive navigation through map-inspired navigational elements and facilitate transitioning between different latent spaces through swappable lenses. We built three VideoMap components to support editors in three common video tasks. In a user study with both professionals and non-professionals, editors found that VideoMap helps reduce grunt work, offers a user-friendly experience, provides an inspirational way of editing, and effectively supports the exploratory nature of video editing. We further demonstrate the versatility of VideoMap by implementing three extended applications. For interactive examples, we invite you to visit our project page: https://chuanenlin.com/videomap.

This paper investigates the challenge of extracting highlight moments from videos. To perform this task, we need to understand what constitutes a highlight for arbitrary video domains while at the same time being able to scale across different domains. Our key insight is that photographs taken by photographers tend to capture the most remarkable or photogenic moments of an activity. Drawing on this insight, we present Videogenic, a technique capable of creating domain-specific highlight videos for a diverse range of domains. In a human evaluation study (N=50), we show that a high-quality photograph collection combined with CLIP-based retrieval (which uses a neural network with semantic knowledge of images) can serve as an excellent prior for finding video highlights. In a within-subjects expert study (N=12), we demonstrate the usefulness of Videogenic in helping video editors create highlight videos with lighter workload, shorter task completion time, and better usability.

Serverless computing offers resource efficiency, cost efficiency, and a "pay-as-you-go" pricing model, which makes it highly attractive to both users and cloud providers. However, serverless computing faces serious cold start problem, especially for deep neural network (DNN) inference, which requires low latency. Existing cold start optimization focuses only on quick container start and fast runtime and library loading. However, DNN application bootstrap (DNN framework load and start, model initialization, model download, deserialization and copy) is the leading factor during the overall cold start time. As the model size grows, the application-level bootstrap becomes more severe.

Diffusion models have shown an impressive ability to model complex data distributions, with several key advantages over GANs, such as stable training, better coverage of the training distribution's modes, and the ability to solve inverse problems without extra training. However, most diffusion models learn the distribution of fixed-resolution images. We propose to learn the distribution of continuous images by training diffusion models on image neural fields, which can be rendered at any resolution, and show its advantages over fixed-resolution models. To achieve this, a key challenge is to obtain a latent space that represents photorealistic image neural fields. We propose a simple and effective method, inspired by several recent techniques but with key changes to make the image neural fields photo-realistic. Our method can be used to convert existing latent diffusion autoencoders into image neural field autoen-coders. We show that image neural field diffusion models can be trained using mixed-resolution image datasets, outperform fixed-resolution diffusion models followed by super-resolution models, and can solve inverse problems with conditions applied at different scales efficiently.

We present Readout Guidance, a method for controlling text-to-image diffusion models with learned signals. Readout Guidance uses readout heads, lightweight networks trained to extract signals from the features of a pre-trained, frozen diffusion model at every timestep. These readouts can encode single-image properties, such as pose, depth, and edges; or higher-order properties that relate multiple images, such as correspondence and appearance similarity. Furthermore, by comparing the readout estimates to a user-defined target, and back-propagating the gradient through the readout head, these estimates can be used to guide the sampling process. Compared to prior methods for conditional generation, Readout Guidance requires significantly fewer added parameters and training samples, and offers a convenient and simple recipe for reproducing different forms of conditional control under a single framework, with a single architecture and sampling procedure. We showcase these benefits in the applications of drag-based manipulation, identity-consistent generation, and spatially aligned control. Project page: https://readout-guidance.github.io.

We present VecFusion, a new neural architecture that can generate vector fonts with varying topological structures and precise control point positions. Our approach is a cascaded diffusion model which consists of a raster diffusion model followed by a vector diffusion model. The raster model generates low-resolution, rasterized fonts with auxiliary control point information, capturing the global style and shape of the font, while the vector model synthesizes vector fonts conditioned on the low-resolution raster fonts from the first stage. To synthesize long and complex curves, our vector diffusion model uses a transformer architecture and a novel vector representation that enables the modeling of diverse vector geometry and the precise prediction of control points. Our experiments show that, in contrast to previous generative models for vector graphics, our new cascaded vector diffusion model generates higher quality vector fonts, with complex structures and diverse styles.

Geometric camera calibration is often required for applications that understand the perspective of the image. We propose Perspective Fields as a representation that models the local perspective properties of an image. Perspective Fields contain per-pixel information about the camera view, parameterized as an Up-vector and a Latitude value. This representation has a number of advantages; it makes minimal assumptions about the camera model and is invariant or equivariant to common image editing operations like cropping, warping, and rotation. It is also more interpretable and aligned with human perception. We train a neural network to predict Perspective Fields and the predicted Perspective Fields can be converted to calibration parameters easily. We demonstrate the robustness of our approach under various scenarios compared with camera calibration-based methods and show example applications in image compositing. Project page: https://jinlinyi.github.io/PerspectiveFields/.

Despite significant progress made in the past few years, challenges remain for depth estimation using a single monocular image. First, it is nontrivial to train a metric-depth prediction model that can generalize well to diverse scenes mainly due to limited training data. Thus, researchers have built large-scale relative depth datasets that are much easier to collect. However, existing relative depth estimation models often fail to recover accurate 3D scene shapes due to the unknown depth shift caused by training with the relative depth data. We tackle this problem here and attempt to estimate accurate scene shapes by training on large-scale relative depth data, and estimating the depth shift. To do so, we propose a two-stage framework that first predicts depth up to an unknown scale and shift from a single monocular image, and then exploits 3D point cloud data to predict the depth shift and the camera's focal length that allow us to recover 3D scene shapes. As the two modules are trained separately, we do not need strictly paired training data. In addition, we propose an image-level normalized regression loss and a normal-based geometry loss to improve training with relative depth annotation. We test our depth model on nine unseen datasets and achieve state-of-the-art performance on zero-shot evaluation. Code is available at: https://github.com/aim-uofa/depth/.

Web tracking by ad networks and other data-driven businesses is often privacy-invasive. Privacy laws, such as the California Consumer Privacy Act, aim to give people more control over their data. In particular, they provide a right to opt out from web tracking via privacy preference signals, notably Global Privacy Control (GPC). GPC holds the promise of enabling people to exercise their opt out rights on the web. Broad adoption of GPC hinges on its usability. In a usability survey we find that 94% of the participants would turn on GPC indicating a need for such efficient and effective opt out mechanism. 81% of the participants in our survey also have a correct understanding of what GPC does ensuring that their intent is accurately represented by their choice. The effectiveness of GPC is dependent on whether websites' GPC compliance can be enforced. A site's GPC compliance can be analyzed based on privacy flags, such as the US Privacy String, which is used on many sites to indicate the opt out status of a web user. Leveraging the US Privacy String for GPC purposes we implement a proof-of-concept browser extension that successfully and correctly analyzes sites' GPC compliance at a rate of 89%. We further implement a web crawler for our browser extension demonstrating that our analysis approach is scalable. We find that many sites do not respect GPC opt out signals despite being legally obligated to do so. Only 54/464 (12%) sites with a US Privacy String opt out users after having received a GPC signal.

We present an interactive system to ease the creation of so-called video doodles - videos on which artists insert hand-drawn animations for entertainment or educational purposes. Video doodles are challenging to create because to be convincing, the inserted drawings must appear as if they were part of the captured scene. In particular, the drawings should undergo tracking, perspective deformations and occlusions as they move with respect to the camera and to other objects in the scene - visual effects that are difficult to reproduce with existing 2D video editing software. Our system supports these effects by relying on planar canvases that users position in a 3D scene reconstructed from the video. Furthermore, we present a custom tracking algorithm that allows users to anchor canvases to static or dynamic objects in the scene, such that the canvases move and rotate to follow the position and direction of these objects. When testing our system, novices could create a variety of short animated clips in a dozen of minutes, while professionals praised its speed and ease of use compared to existing tools.

We introduce 3INGAN, an unconditional 3D generative model trained from 2D images of a single self-similar 3D scene. Such a model can be used to produce 3D “remixes” of a given scene, by mapping spatial latent codes into a 3D volumetric representation, which can subsequently be rendered from arbitrary views using physically based volume rendering. By construction, the generated scenes remain view-consistent across arbitrary camera configurations, without any flickering or spatio-temporal artifacts. During training, we employ a combination of 2D, obtained through differentiable volume tracing, and 3D Generative Adversarial Network (GAN) losses, across multiple scales, enforcing realism on both its 2D renderings and its 3D structure. We show results on semi-stochastic scenes of varying scale and complexity, obtained from real and synthetic sources. We demonstrate, for the first time, the feasibility of learning plausible view-consistent 3D scene variations from a single exemplar scene and provide qualitative and quantitative comparisons against two recent related methods. Code and data for the paper are available at https://geometry.cs.ucl.ac.uk/group_website/projects/2022/3inGAN/.

We introduce an approach for selecting objects in neural volumetric 3D representations, such as multi-plane images (MPI) and neural radiance fields (NeRF). Our approach takes a set of foreground and background 2D user scribbles in one view and automatically estimates a 3D segmentation of the desired object, which can be rendered into novel views. To achieve this result, we propose a novel voxel feature embedding that incorporates the neural volumetric 3D representation and multi-view image features from all input views. To evaluate our approach, we introduce a new dataset of human-provided segmentation masks for depicted objects in real-world multi-view scene captures. We show that our approach out-performs strong baselines, including 2D segmentation and 3D segmentation approaches adapted to our task.

We introduce CharacterGAN, a generative model that can be trained on only a few samples (8 – 15) of a given character. Our model generates novel poses based on keypoint locations, which can be modified in real time while providing interactive feedback, allowing for intuitive reposing and animation. Since we only have very limited training samples, one of the key challenges lies in how to address (dis)occlusions, e.g. when a hand moves behind or in front of a body. To address this, we introduce a novel layering approach which explicitly splits the input keypoints into different layers which are processed independently. These layers represent different parts of the character and provide a strong implicit bias that helps to obtain realistic results even with strong (dis)occlusions. To combine the features of individual layers we use an adaptive scaling approach conditioned on all keypoints. Finally, we introduce a mask connectivity constraint to reduce distortion artifacts that occur with extreme out-of-distribution poses at test time. We show that our approach outperforms recent baselines and creates realistic animations for diverse characters. We also show that our model can handle discrete state changes, for example a profile facing left or right, that the different layers do indeed learn features specific for the respective keypoints in those layers, and that our model scales to larger datasets when more data is available. Code is available at https://github.com/tohinz/CharacterGAN.

Despite significant progress in monocular depth estimation in the wild, recent state-of-the-art methods cannot be used to recover accurate 3D scene shape due to an unknown depth shift induced by shift-invariant reconstruction losses used in mixed-data depth prediction training, and possible unknown camera focal length. We investigate this problem in detail, and propose a two-stage framework that first predicts depth up to an unknown scale and shift from a single monocular image, and then use 3D point cloud encoders to predict the missing depth shift and focal length that allow us to recover a realistic 3D scene shape. In addition, we propose an image-level normalized regression loss and a normal-based geometry loss to enhance depth prediction models trained on mixed datasets. We test our depth model on nine unseen datasets and achieve state-of-the-art performance on zero-shot dataset generalization. Code is available at: https://git.io/Depth

We present a method to perform novel view and time synthesis of dynamic scenes, requiring only a monocular video with known camera poses as input. To do this, we introduce Neural Scene Flow Fields, a new representation that models the dynamic scene as a time-variant continuous function of appearance, geometry, and 3D scene motion. Our representation is optimized through a neural network to fit the observed input views. We show that our representation can be used for varieties of in-the-wild scenes, including thin structures, view-dependent effects, and complex degrees of motion. We conduct a number of experiments that demonstrate our approach significantly outperforms recent monocular view synthesis methods, and show qualitative results of space-time view synthesis on a variety of real-world videos.

We present a data-driven approach to automate audio signal processing by incorporating stateful third-party, audio effects as layers within a deep neural network. We then train a deep encoder to analyze input audio and control effect parameters to perform the desired signal manipulation, requiring only input-target paired audio data as supervision. To train our network with non-differentiable black-box effects layers, we use a fast, parallel stochastic gradient approximation scheme within a standard auto differentiation graph, yielding efficient end-to-end backpropagation. We demonstrate the power of our approach with three separate automatic audio production applications: tube amplifier emulation, automatic removal of breaths and pops from voice recordings, and automatic music mastering. We validate our results with a subjective listening test, showing our approach not only can enable new automatic audio effects tasks, but can yield results comparable to a specialized, state-of-the-art commercial solution for music mastering.

We introduce an approach for computing a 3D mesh from one or more views of an object by establishing dense correspondences between pixels in the views and 3D locations on a learnable parameterized surface. We propose a multi-view shape encoder that can be jointly trained with the AtlasNet surface parameterization. The shape is further refined using a novel geometric cycle-consistency loss between the learnable parameterized surface and input views. We demonstrate the efficacy of our approach on the ShapeNet-COCO dataset.

Recently there has been an interest in the potential of learning generative models from a single image, as opposed to from a large dataset. This task is of significance, as it means that generative models can be used in domains where collecting a large dataset is not feasible. However, training a model capable of generating realistic images from only a single sample is a difficult problem. In this work, we conduct a number of experiments to understand the challenges of training these methods and propose some best practices that we found allowed us to generate improved results over previous work. One key piece is that, unlike prior single image generation methods, we concurrently train several stages in a sequential multi-stage manner, allowing us to learn models with fewer stages of increasing image resolution. Compared to a recent state of the art baseline, our model is up to six times faster to train, has fewer parameters, and can better capture the global structure of images.

Video frame interpolation, the synthesis of novel views in time, is an increasingly popular research direction with many new papers further advancing the state of the art. But as each new method comes with a host of variables that affect the interpolation quality, it can be hard to tell what is actually important for this task. In this work, we show, somewhat surprisingly, that it is possible to achieve near state-of-the-art results with an older, simpler approach, namely adaptive separable convolutions, by a subtle set of low level improvements. In doing so, we propose a number of intuitive but effective techniques to improve the frame interpolation quality, which also have the potential to other related applications of adaptive convolutions such as burst image denoising, joint image filtering, or video prediction.

We present a method that decomposes, and "unwraps", an input video into a set of layered 2D atlases , each providing a unified representation of the appearance of an object (or background) over the video. For each pixel in the video, our method estimates its corresponding 2D coordinate in each of the atlases, giving us a consistent parameterization of the video, along with an associated alpha (opacity) value. Importantly, we design our atlases to be interpretable and semantic, which facilitates easy and intuitive editing in the atlas domain, with minimal manual work required. Edits applied to a single 2D atlas (or input video frame) are automatically and consistently mapped back to the original video frames, while preserving occlusions, deformation, and other complex scene effects such as shadows and reflections. Our method employs a coordinate-based Multilayer Perceptron (MLP) representation for mappings, atlases, and alphas, which are jointly optimized on a per-video basis, using a combination of video reconstruction and regularization losses. By operating purely in 2D, our method does not require any prior 3D knowledge about scene geometry or camera poses, and can handle complex dynamic real world videos. We demonstrate various video editing applications, including texture mapping, video style transfer, image-to-video texture transfer, and segmentation/labeling propagation, all automatically produced by editing a single 2D atlas image.

We present TDNet, a temporally distributed network designed for fast and accurate video semantic segmentation. We observe that features extracted from a certain high-level layer of a deep CNN can be approximated by composing features extracted from several shallower sub-networks. Leveraging the inherent temporal continuity in videos, we distribute these sub-networks over sequential frames. Therefore, at each time step, we only need to perform a lightweight computation to extract a sub-features group from a single sub-network. The full features used for segmentation are then recomposed by application of a novel attention propagation module that compensates for geometry deformation between frames. A grouped knowledge distillation loss is also introduced to further improve the representation power at both full and sub-feature levels. Experiments on Cityscapes, CamVid, and NYUD-v2 demonstrate that our method achieves state-of-the-art accuracy with significantly faster speed and lower latency.

While Generative Adversarial Networks (GANs) have seen huge successes in image synthesis tasks, they are notoriously difficult to adapt to different datasets, in part due to instability during training and sensitivity to hyperparameters. One commonly accepted reason for this instability is that gradients passing from the discriminator to the generator become uninformative when there isn't enough overlap in the supports of the real and fake distributions. In this work, we propose the Multi-Scale Gradient Generative Adversarial Network (MSG-GAN), a simple but effective technique for addressing this by allowing the flow of gradients from the discriminator to the generator at multiple scales. This technique provides a stable approach for high resolution image synthesis, and serves as an alternative to the commonly used progressive growing technique. We show that MSG-GAN converges stably on a variety of image datasets of different sizes, resolutions and domains, as well as different types of loss functions and architectures, all with the same set of fixed hyperparameters. When compared to state-of-the-art GANs, our approach matches or exceeds the performance in most of the cases we tried.

In this work we ask whether it is possible to create a "universal" detector for telling apart real images from these generated by a CNN, regardless of architecture or dataset used. To test this, we collect a dataset consisting of fake images generated by 11 different CNN-based image generator models, chosen to span the space of commonly used architectures today (ProGAN, StyleGAN, BigGAN, CycleGAN, StarGAN, GauGAN, DeepFakes, cascaded refinement networks, implicit maximum likelihood estimation, second-order attention super-resolution, seeing-in-the-dark). We demonstrate that, with careful pre- and post-processing and data augmentation, a standard image classifier trained on only one specific CNN generator (ProGAN) is able to generalize surprisingly well to unseen architectures, datasets, and training methods (including the just released StyleGAN2). Our findings suggest the intriguing possibility that today's CNN-generated images share some common systematic flaws, preventing them from achieving realistic image synthesis.

Monocular depth estimation is an ill-posed problem, and as such critically relies on scene priors and semantics. Due to its complexity, we propose a deep neural network model based on a semantic divide-and-conquer approach. Our model decomposes a scene into semantic segments, such as object instances and background stuff classes, and then predicts a scale and shift invariant depth map for each semantic segment in a canonical space. Semantic segments of the same category share the same depth decoder, so the global depth prediction task is decomposed into a series of category-specific ones, which are simpler to learn and easier to generalize to new scene types. Finally, our model stitches each local depth segment by predicting its scale and shift based on the global context of the image. The model is trained end-to-end using a multi-task loss for panoptic segmentation and depth prediction, and is therefore able to leverage large-scale panoptic segmentation datasets to boost its semantic understanding. We validate the effectiveness of our approach and show state-of-the-art performance on three benchmark datasets.

Single image depth prediction is a challenging task due to its ill-posed nature and challenges with capturing ground truth for supervision. Large-scale disparity data generated from stereo photos and 3D videos is a promising source of supervision, however, such disparity data can only approximate the inverse ground truth depth up to an affine transformation. To more effectively learn from such pseudo-depth data, we propose to use a simple pair-wise ranking loss with a novel sampling strategy. Instead of randomly sampling point pairs, we guide the sampling to better characterize structure of important regions based on the low-level edge maps and high-level object instance masks. We show that the pair-wise ranking loss, combined with our structure-guided sampling strategies, can significantly improve the quality of depth map prediction. In addition, we introduce a new relative depth dataset of about 21K diverse high-resolution web stereo photos to enhance the generalization ability of our model. In experiments, we conduct cross-dataset evaluation on six benchmark datasets and show that our method consistently improves over the baselines, leading to superior quantitative and qualitative results.

Deep generative models have become increasingly effective at producing realistic images from randomly sampled seeds, but using such models for controllable manipulation of existing images remains challenging. We propose the Swapping Autoencoder, a deep model designed specifically for image manipulation, rather than random sampling. The key idea is to encode an image into two independent components and enforce that any swapped combination maps to a realistic image. In particular, we encourage the components to represent structure and texture, by enforcing one component to encode co-occurrent patch statistics across different parts of the image. As our method is trained with an encoder, finding the latent codes for a new input image becomes trivial, rather than cumbersome. As a result, our method enables us to manipulate real input images in various ways, including texture swapping, local and global editing, and latent code vector arithmetic. Experiments on multiple datasets show that our model produces better results and is substantially more efficient compared to recent generative models.

We present a fully data-driven method to compute depth from diverse monocular video sequences that contain large amounts of non-rigid objects, e.g., people. In order to learn reconstruction cues for non-rigid scenes, we introduce a new dataset consisting of stereo videos scraped in-the-wild. This dataset has a wide variety of scene types, and features large amounts of nonrigid objects, especially people. From this, we compute disparity maps to be used as supervision to train our approach. We propose a loss function that allows us to generate a depth prediction even with unknown camera intrinsics and stereo baselines in the dataset. We validate the use of large amounts of Internet video by evaluating our method on existing video datasets with depth supervision, including SINTEL, and KITTI, and show that our approach generalizes better to natural scenes.

When watching how-to videos related to physical tasks, users' hands are often occupied by the task, making voice input a natural fit. To better understand the design space of voice interactions for how-to video navigation, we conducted three think-aloud studies using: 1) a traditional video interface, 2) a research probe providing a voice controlled video interface, and 3) a wizard-of-oz interface. From the studies, we distill seven navigation objectives and their underlying intents: pace control pause, content alignment pause, video control pause, reference jump, replay jump, skip jump, and peek jump. Our analysis found that users' navigation objectives and intents affect the choice of referent type and referencing approach in command utterances. Based on our findings, we recommend to 1) support conversational strategies like sequence expansions and command queues, 2) allow users to identify and refine their navigation objectives explicitly, and 3) support the seven interaction intents.

In video production, inserting B-roll is a widely used technique to enrich the story and make a video more engaging. However, determining the right content and positions of B-roll and actually inserting it within the main footage can be challenging, and novice producers often struggle to get both timing and content right. We present B-Script, a system that supports B-roll video editing via interactive transcripts. B-Script has a built-in recommendation system trained on expert-annotated data, recommending users B-roll position and content. To evaluate the system, we conducted a within-subject user study with 110 participants, and compared three interface variations: a timeline-based editor, a transcript-based editor, and a transcript-based editor with recommendations. Users found it easier and were faster to insert B-roll using the transcript-based interface, and they created more engaging videos when recommendations were provided.

In this paper, we address the problem of 3D object mesh reconstruction from RGB videos. Our approach combines the best of multi-view geometric and data-driven methods for 3D reconstruction by optimizing object meshes for multi-view photometric consistency while constraining mesh deformations with a shape prior. We pose this as a piecewise image alignment problem for each mesh face projection. Our approach allows us to update shape parameters from the photometric error without any depth or mask information. Moreover, we show how to avoid a degeneracy of zero photometric gradients via rasterizing from a virtual viewpoint. We demonstrate 3D object mesh reconstruction results from both synthetic and real-world videos with our photometric mesh optimization, which is unachievable with either naive mesh generation networks or traditional pipelines of surface reconstruction without heavy manual post-processing.

We propose an interactive GAN-based sketch-to-image translation method that helps novice users easily create images of simple objects. The user starts with a sparse sketch and a desired object category, and the network then recommends its plausible completion(s) and shows a corresponding synthesized image. This enables a feedback loop, where the user can edit the sketch based on the network's recommendations, while the network is able to better synthesize the image that the user might have in mind. In order to use a single model for a wide array of object classes, we introduce a gating-based approach for class conditioning, which allows us to generate distinct classes without feature mixing, from a single generator network.

Most malicious photo manipulations are created using standard image editing tools, such as Adobe Photoshop. We present a method for detecting one very popular Photoshop manipulation -- image warping applied to human faces -- using a model trained entirely using fake images that were automatically generated by scripting Photoshop itself. We show that our model outperforms humans at the task of recognizing manipulated images, can predict the specific location of edits, and in some cases can be used to "undo" a manipulation to reconstruct the original, unedited image. We demonstrate that the system can be successfully applied to artist-created image manipulations.

Three-hundred-sixty-degree (360°) video provides an immersive experience for viewers, allowing them to freely explore the world by turning their head. However, creating high-quality 360° video content can be challenging, as viewers may miss important events by looking in the wrong direction, or they may see things that ruin the immersion, such as stitching artifacts and the film crew. We take advantage of the fact that not all directions are equally likely to be observed; most viewers are more likely to see content located at “true north,” i.e., in front of them, due to ergonomic constraints. We therefore propose 360° video direction, where the video is jointly optimized to orient important events to the front of the viewer and visual clutter behind them, while producing smooth camera motion. Unlike traditional video, viewers can still explore the space as desired, but with the knowledge that the most important content is likely to be in front of them. Constraints can be user guided, either added directly on the equirectangular projection or by recording “guidance” viewing directions while watching the video in a VR headset or automatically computed, such as via visual saliency or forward-motion direction. To accomplish this, we propose a new motion estimation technique specifically designed for 360° video that outperforms the commonly used five-point algorithm on wide-angle video. We additionally formulate the direction problem as an optimization where a novel parametrization of spherical warping allows us to correct for some degree of parallax effects. We compare our approach to recent methods that address stabilization-only and converting 360° video to narrow field-of-view video. Our pipeline can also enable the viewing of wide-angle non-360° footage in a spherical 360° space, giving an immersive “virtual cinema” experience for a wide range of existing content filmed with first-person cameras.

We address the problem of finding realistic geometric corrections to a foreground object such that it appears natural when composited into a background image. To achieve this, we propose a novel Generative Adversarial Network (GAN) architecture that utilizes Spatial Transformer Networks (STNs) as the generator, which we call Spatial Transformer GANs (ST-GANs). ST-GANs seek image realism by operating in the geometric warp parameter space. In particular, we exploit an iterative STN warping scheme and propose a sequential training strategy that achieves better results compared to naive training of a single generator. One of the key advantages of ST-GAN is its applicability to high-resolution images indirectly since the predicted warp parameters are transferable between reference frames. We demonstrate our approach in two applications: (1) visualizing how indoor furniture (e.g. from product images) might be perceived in a room, (2) hallucinating how accessories like glasses would look when matched with real portraits.

While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called "perceptual losses"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations.

Localizing moments in a longer video via natural language queries is a new, challenging task at the intersection of language and video understanding.Though moment localization with natural language is similar to other language and vision tasks like natural language object retrieval in images, moment localization offers an interesting opportunity to model temporal dependencies and reasoning in text.We propose a new model that explicitly reasons about different temporal segments in a video, and shows that temporal context is important for localizing phrases which include temporal language.To benchmark whether our model, and other recent video localization models, can effectively reason about temporal language, we collect the novel TEMPOral reasoning in video and language (TEMPO) dataset.Our dataset consists of two parts: a dataset with real videos and template sentences (TEMPO -Template Language) which allows for controlled studies on temporal language, and a human language dataset which consists of temporal sentences annotated by humans (TEMPO -Human Language).

We introduce an approach to convert mono audio recorded by a 360° video camera into spatial audio, a representation of the distribution of sound over the full viewing sphere. Spatial audio is an important component of immersive 360° video viewing, but spatial audio microphones are still rare in current 360° video production. Our system consists of end-to-end trainable neural networks that separate individual sound sources and localize them on the viewing sphere, conditioned on multi-modal analysis from the audio and 360° video frames. We introduce several datasets, including one filmed ourselves, and one collected in-the-wild from YouTube, consisting of 360° videos uploaded with spatial audio. During training, ground truth spatial audio serves as self-supervision and a mixed down mono track forms the input to our network. Using our approach we show that it is possible to infer the spatial localization of sounds based only on a synchronized 360° video and the mono audio track.

Abstract In this work, we present a method to vectorize raster images of line art. Inverting the rasterization procedure is inherently ill‐conditioned, as there exist many possible vector images that could yield the same raster image. However, not all of these vector images are equally useful to the user, especially if performing further edits is desired. We therefore define the problem of computing an instance segmentation of the most likely set of paths that could have created the raster image. Once the segmentation is computed, we use existing vectorization approaches to vectorize each path, and then combine all paths into the final output vector image. To determine which set of paths is most likely, we train a pair of neural networks to provide semantic clues that help resolve ambiguities at intersection and overlap regions. These predictions are made considering the full context of the image, and are then globally combined by solving a Markov Random Field (MRF). We demonstrate the flexibility of our method by generating results on character datasets, a synthetic random line dataset, and a dataset composed of human drawn sketches. For all cases, our system accurately recovers paths that adhere to the semantics of the drawings.

We aim to generate high resolution shallow depth-of-field (DoF) images from a single all-in-focus image with controllable focal distance and aperture size. To achieve this, we propose a novel neural network model comprised of a depth prediction module, a lens blur module, and a guided upsampling module. All modules are differentiable and are learned from data. To train our depth prediction module, we collect a dataset of 2462 RGB-D images captured by mobile phones with a dual-lens camera, and use existing segmentation datasets to improve border prediction. We further leverage a synthetic dataset with known depth to supervise the lens blur and guided upsampling modules. The effectiveness of our system and training strategies are verified in the experiments. Our method can generate high-quality shallow DoF images at high resolution, and produces significantly fewer artifacts than the baselines and existing solutions for single image shallow DoF synthesis. Compared with the iPhone portrait mode, which is a state-of-the-art shallow DoF solution based on a dual-lens depth camera, our method generates comparable results, while allowing for greater flexibility to choose focal points and aperture size, and is not limited to one capture setup.

Many monocular visual SLAM algorithms are derived from incremental structure-from-motion (SfM) methods. This work proposes a novel monocular SLAM method which integrates recent advances made in global SfM. In particular, we present two main contributions to visual SLAM. First, we solve the visual odometry problem by a novel rank-1 matrix factorization technique which is more robust to the errors in map initialization. Second, we adopt a recent global SfM method for the pose-graph optimization, which leads to a multi-stage linear formulation and enables L1 optimization for better robustness to false loops. The combination of these two approaches generates more robust reconstruction and is significantly faster (4X) than recent state-of-the-art SLAM systems. We also present a new dataset recorded with ground truth camera motion in a Vicon motion capture room, and compare our method to prior systems on it and established benchmark datasets.

Motion blur from camera shake is a major problem in videos captured by hand-held devices. Unlike single-image deblurring, video-based approaches can take advantage of the abundant information that exists across neighboring frames. As a result the best performing methods rely on the alignment of nearby frames. However, aligning images is a computationally expensive and fragile procedure, and methods that aggregate information must therefore be able to identify which regions have been accurately aligned and which have not, a task that requires high level scene understanding. In this work, we introduce a deep learning solution to video deblurring, where a CNN is trained end-to-end to learn how to accumulate information across frames. To train this network, we collected a dataset of real videos recorded with a high frame rate camera, which we use to generate synthetic motion blur for supervision. We show that the features learned from this dataset extend to deblurring motion blur that arises due to camera shake in a wide range of videos, and compare the quality of results to a number of other baselines.

We consider retrieving a specific temporal segment, or moment, from a video given a natural language text description. Methods designed to retrieve whole video clips with natural language determine what occurs in a video but not when. To address this issue, we propose the Moment Context Network (MCN) which effectively localizes natural language queries in videos by integrating local and global video features over time. A key obstacle to training our MCN model is that current video datasets do not include pairs of localized video segments and referring expressions, or text descriptions which uniquely identify a corresponding moment. Therefore, we collect the Distinct Describable Moments (DiDeMo) dataset which consists of over 10,000 unedited, personal videos in diverse visual settings with pairs of localized video segments and referring expressions. We demonstrate that MCN outperforms several baseline methods and believe that our initial results together with the release of DiDeMo will inspire further research on localizing video moments with natural language.

Many image-to-image translation problems are ambiguous, as a single input image may correspond to multiple possible outputs. In this work, we aim to model a distribution of possible outputs in a conditional generative modeling setting. The ambiguity of the mapping is distilled in a low-dimensional latent vector, which can be randomly sampled at test time. A generator learns to map the given input, combined with this latent code, to the output. We explicitly encourage the connection between output and the latent code to be invertible. This helps prevent a many-to-one mapping from the latent code to the output during training, also known as the problem of mode collapse, and produces more diverse results. We explore several variants of this approach by employing different training objectives, network architectures, and methods of injecting the latent code. Our proposed method encourages bijective consistency between the latent encoding and output modes. We present a systematic comparison of our method and other variants on both perceptual realism and diversity.

Visual formats have advanced beyond single-view images and videos: 3D movies are commonplace, researchers have developed multi-view navigation systems, and VR is helping to push light field cameras to mass market. However, editing tools for these media are still nascent, and even simple filtering operations like color correction or stylization are problematic: naively applying image filters per frame or per view rarely produces satisfying results due to time and space inconsistencies. Our method preserves and stabilizes filter effects while being agnostic to the inner working of the filter. It captures filter effects in the gradient domain, then uses input frame gradients as a reference to impose temporal and spatial consistency. Our least-squares formulation adds minimal overhead compared to naive data processing. Further, when filter cost is high, we introduce a filter transfer strategy that reduces the number of per-frame filtering computations by an order of magnitude, with only a small reduction in visual quality. We demonstrate our algorithm on several camera array formats including stereo videos, light fields, and wide baselines.

Time slice photography is a popular effect that visualizes the passing of time by aligning and stitching multiple images capturing the same scene at different times together into a single image. Extending this effect to video is a difficult problem, and one where existing solutions have only had limited success. In this paper, we propose an easy-to-use and robust system for creating time slice videos from a wide variety of consumer videos. The main technical challenge we address is how to align videos taken at different times with substantially different appearances, in the presence of moving objects and moving cameras with slightly different trajectories. To achieve a temporally stable alignment, we perform a mixed 2D-3D alignment, where a rough 3D reconstruction is used to generate sparse constraints that are integrated into a pixelwise 2D registration. We apply our method to a number of challenging scenarios, and show that we can achieve a higher quality registration than prior work. We propose a 3D user interface that allows the user to easily specify how multiple videos should be composited in space and time. Finally, we show that our alignment method can be applied in more general video editing and compositing tasks, such as object removal.

We present a novel approach to facilitate the creation of stylized 2D rigid body animations. Our approach can handle multiple rigid objects following complex physically-simulated trajectories with collisions, while retaining a unique artistic style directly specified by the user. Starting with an existing target animation (e.g., produced by a physical simulation engine) an artist interactively draws over a sparse set of frames, and the desired appearance and motion stylization is automatically propagated to the rest of the sequence. The stylization process may also be performed in an off-line batch process from a small set of drawn sequences. To achieve these goals, we combine parametric deformation synthesis that generalizes and reuses hand-drawn exemplars, with non-parametric techniques that enhance the hand-drawn appearance of the synthesized sequence. We demonstrate the potential of our method on various complex rigid body animations which are created with an expressive hand-drawn look using notably less manual interventions as compared to traditional techniques.

In this work, we propose a novel approach to video segmentation that operates in bilateral space. We design a new energy on the vertices of a regularly sampled spatiotemporal bilateral grid, which can be solved efficiently using a standard graph cut label assignment. Using a bilateral formulation, the energy that we minimize implicitly approximates long-range, spatio-temporal connections between pixels while still containing only a small number of variables and only local graph edges. We compare to a number of recent methods, and show that our approach achieves state-of-the-art results on multiple benchmarks in a fraction of the runtime. Furthermore, our method scales linearly with image size, allowing for interactive feedback on real-world high resolution video.

We present a new approach for efficient approximate nearest neighbor (ANN) search in high dimensional spaces, extending the idea of Product Quantization. We propose a two level product and vector quantization tree that reduces the number of vector comparisons required during tree traversal. Our approach also includes a novel highly parallelizable re-ranking method for candidate vectors by efficiently reusing already computed intermediate values. Due to its small memory footprint during traversal the method lends itself to an efficient, parallel GPU implementation. This Product Quantization Tree (PQT) approach significantly outperforms recent state of the art methods for high dimensional nearest neighbor queries on standard reference datasets. Ours is the first work that demonstrates GPU performance superior to CPU performance on high dimensional, large scale ANN problems in time-critical real-world applications, like loop-closing in videos.

Precise object segmentation in image data is a fundamental problem with various applications, including 3D object reconstruction. We present an efficient algorithm to automatically segment a static foreground object from highly cluttered background in light fields. A key insight and contribution of our article is that a significant increase of the available input data can enable the design of novel, highly efficient approaches. In particular, the central idea of our method is to exploit high spatio-angular sampling on the order of thousands of input frames, for example, captured as a hand-held video, such that new structures are revealed due to the increased coherence in the data. We first show how purely local gradient information contained in slices of such a dense light field can be combined with information about the camera trajectory to make efficient estimates of the foreground and background. These estimates are then propagated to textureless regions using edge-aware filtering in the epipolar volume. Finally, we enforce global consistency in a gathering step to derive a precise object segmentation in both 2D and 3D space, which captures fine geometric details even in very cluttered scenes. The design of each of these steps is motivated by efficiency and scalability, allowing us to handle large, real-world video datasets on a standard desktop computer. We demonstrate how the results of our method can be used for considerably improving the speed and quality of image-based 3D reconstruction algorithms, and we compare our results to state-of-the-art segmentation and multiview stereo methods.

Standard approaches to computing interpolated (in-between) frames in a video sequence require accurate pixel correspondences between images e.g. using optical flow. We present an efficient alternative by leveraging recent developments in phase-based methods that represent motion in the phase shift of individual pixels. This concept allows in-between images to be generated by simple per-pixel phase modification, without the need for any form of explicit correspondence estimation. Up until now, such methods have been limited in the range of motion that can be interpolated, which fundamentally restricts their usefulness. In order to reduce these limitations, we introduce a novel, bounded phase shift correction method that combines phase information across the levels of a multi-scale pyramid. Additionally, we propose extensions for phase-based image synthesis that yield smoother transitions between the interpolated images. Our approach avoids expensive global optimization typical of optical flow methods, and is both simple to implement and easy to parallelize. This allows us to interpolate frames at a fraction of the computational cost of traditional optical flow-based solutions, while achieving similar quality and in some cases even superior results. Our method fails gracefully in difficult interpolation settings, e.g., significant appearance changes, where flow-based methods often introduce serious visual artifacts. Due to its efficiency, our method is especially well suited for frame interpolation and retiming of high resolution, high frame rate video.

Videos consisting of thousands of high resolution frames are challenging for existing structure from motion (SfM) and simultaneous-localization and mapping (SLAM) techniques. We present a new approach for simultaneously computing extrinsic camera poses and 3D scene structure that is capable of handling such large volumes of image data. The key insight behind this paper is to effectively exploit coherence in densely sampled video input. Our technical contributions include robust tracking and selection of confident video frames, a novel window bundle adjustment, frame-to-structure verification for globally consistent reconstructions with multi-loop closing, and utilizing efficient global linear camera pose estimation in order to link both consecutive and distant bundle adjustment windows. To our knowledge we describe the first system that is capable of handling high resolution, high frame-rate video data with close to real-time performance. In addition, our approach can robustly integrate data from different video sequences, allowing multiple video streams to be simultaneously calibrated in an efficient and globally optimal way. We demonstrate high quality alignment on large scale challenging datasets, e.g., 2-20 megapixel resolution at frame rates of 25-120 Hz with thousands of frames.

We present a method to continuously blend between multiple facial performances of an actor, which can contain different facial expressions or emotional states. As an example, given sad and angry video takes of a scene, our method empowers the movie director to specify arbitrary weighted combinations and smooth transitions between the two takes in post-production. Our contributions include (1) a robust nonlinear audio-visual synchronization technique that exploits complementary properties of audio and visual cues to automatically determine robust, dense spatiotemporal correspondences between takes, and (2) a seamless facial blending approach that provides the director full control to interpolate timing, facial expression, and local appearance, in order to generate novel performances after filming. In contrast to most previous works, our approach operates entirely in image space, avoiding the need of 3D facial reconstruction. We demonstrate that our method can synthesize visually believable performances with applications in emotion transition, performance correction, and timing control.

We present a novel approach to video segmentation using multiple object proposals. The problem is formulated as a minimization of a novel energy function defined over a fully connected graph of object proposals. Our model combines appearance with long-range point tracks, which is key to ensure robustness with respect to fast motion and occlusions over longer video sequences. As opposed to previous approaches based on object proposals, we do not seek the best per-frame object hypotheses to perform the segmentation. Instead, we combine multiple, potentially imperfect proposals to improve overall segmentation accuracy and ensure robustness to outliers. Overall, the basic algorithm consists of three steps. First, we generate a very large number of object proposals for each video frame using existing techniques. Next, we perform an SVM-based pruning step to retain only high quality proposals with sufficiently discriminative power. Finally, we determine the fore-and background classification by solving for the maximum a posteriori of a fully connected conditional random field, defined using our novel energy function. Experimental results on a well established dataset demonstrate that our method compares favorably to several recent state-of-the-art approaches.

Abstract We describe an algorithm for generating panoramic video from unstructured camera arrays. Artifact‐free panorama stitching is impeded by parallax between input views. Common strategies such as multi‐level blending or minimum energy seams produce seamless results on quasi‐static input. However, on video input these approaches introduce noticeable visual artifacts due to lack of global temporal and spatial coherence. In this paper we extend the basic concept of local warping for parallax removal. Firstly, we introduce an error measure with increased sensitivity to stitching artifacts in regions with pronounced structure. Using this measure, our method efficiently finds an optimal ordering of pair‐wise warps for robust stitching with minimal parallax artifacts. Weighted extrapolation of warps in non‐overlap regions ensures temporal stability, while at the same time avoiding visual discontinuities around transitions between views. Remaining global deformation introduced by the warps is spread over the entire panorama domain using constrained relaxation, while staying as close as possible to the original input views. In combination, these contributions form the first system for spatiotemporally stable panoramic video stitching from unstructured camera array input.

Abstract Renderings of animation sequences with physics‐based Monte Carlo light transport simulations are exceedingly costly to generate frame‐by‐frame, yet much of this computation is highly redundant due to the strong coherence in space, time and among samples. A promising approach pursued in prior work entails subsampling the sequence in space, time, and number of samples, followed by image‐based spatio‐temporal upsampling and denoising. These methods can provide significant performance gains, though major issues remain: firstly, in a multiple scattering simulation, the final pixel color is the composite of many different light transport phenomena, and this conflicting information causes artifacts in image‐based methods. Secondly, motion vectors are needed to establish correspondence between the pixels in different frames, but it is unclear how to obtain them for most kinds of light paths (e.g. an object seen through a curved glass panel). To reduce these ambiguities, we propose a general decomposition framework, where the final pixel color is separated into components corresponding to disjoint subsets of the space of light paths. Each component is accompanied by motion vectors and other auxiliary features such as reflectance and surface normals. The motion vectors of specular paths are computed using a temporal extension of manifold exploration and the remaining components use a specialized variant of optical flow. Our experiments show that this decomposition leads to significant improvements in three image‐based applications: denoising, spatial upsampling, and temporal interpolation.

Many compelling video processing effects can be achieved if per-pixel depth information and 3D camera calibrations are known. However, the success of such methods is highly dependent on the accuracy of this "scene-space" information. We present a novel, sampling-based framework for processing video that enables high-quality scene-space video effects in the presence of inevitable errors in depth and camera pose estimation. Instead of trying to improve the explicit 3D scene representation, the key idea of our method is to exploit the high redundancy of approximate scene information that arises due to most scene points being visible multiple times across many frames of video. Based on this observation, we propose a novel pixel gathering and filtering approach. The gathering step is general and collects pixel samples in scene-space, while the filtering step is application-specific and computes a desired output video from the gathered sample sets. Our approach is easily parallelizable and has been implemented on GPU, allowing us to take full advantage of large volumes of video data and facilitating practical runtimes on HD video using a standard desktop computer. Our generic scene-space formulation is able to comprehensively describe a multitude of video processing applications such as denoising, deblurring, super resolution, object removal, computational shutter functions, and other scene-space camera effects. We present results for various casually captured, hand-held, moving, compressed, monocular videos depicting challenging scenes recorded in uncontrolled environments.

We propose an original 2D agent-based model of biological "autopoiesis", the process by which a cell creates and continuously regenerates itself, considered one of the defining characteristics of life.In the space of our simulations, the positions of free molecules are continuous, and polymerized membrane components are regularly arranged (plus noise).While autopoiesis commonly refers to the self-driven maintenance of a system, we also follow Varela's historical study of emergence, and show that the same model can account for both selfperpetuation and self-formation-a step toward uniting the three main perspectives on life: origins, autopoiesis, and replication.Exploring different initial and environmental conditions, we observe that destructive reactions are important for the survival of our autopoietic system, and evaluate their impact on its lifespan.The tendency of cells to form spurious outgrowths is counteracted by moderate decay of the membrane.

Free viewpoint video enables interactive viewpoint selection in real world scenes, which is attractive for many applications such as sports visualization. Multi-camera registration is one of the difficult tasks in such systems. We introduce the concept of a static high resolution master camera for improved long-term multiview alignment. All broadcast cameras are aligned to a common reference. Our approach builds on frame-to-frame alignment, extended into a recursive long-term estimation process, which is shown to be accurate, robust and stable over long sequences.

Water painting is an art, where the result and experience strongly depend on the process of creating it. Color is injected by the artist in a controlled way and diffuses into a final state. We present a system to simulate such a process, which builds on existing colorization approaches. The geodesic distance builds the mathematical foundation for our animated diffusion. We add a time-dependent weight function to generate a diffusion-like spreading effect. Tiled approximation is used to achieve interactive rates on modern mobile devices. The result is a colorization framework simulating water painting that allows giving real-time feedback on touch events with the limited hardware resources of a tablet computer.

Aligning video is a fundamental task in computer graphics and vision, required for a wide range of applications. We present aninteractivemethod for computing optimal nonlinear temporal video alignments of an arbitrary number of videos. We first derive a robust approximation of alignment quality between pairs of clips, computed as a weighted histogram of feature matches. We then find optimal temporal mappings (constituting frame correspondences) using a graph-based approach that allows for very efficient evaluation with artist constraints. This enables an enhancement to the "snapping" interface in video editing tools, where videos in a time-line are now able snap to one another when dragged by an artist based on theircontent, rather than simply start-and-end times. The pairwise snapping is then generalized to multiple clips, achieving a globally optimal temporal synchronization that automatically arranges a series of clips filmed at different times into a single consistent time frame. When followed by a simple spatial registration, we achieve high quality spatiotemporal video alignments at a fraction of the computational complexity compared to previous methods. Assisted temporal alignment is a degree of freedom that has been largely unexplored, but is an important task in video editing. Our approach is simple to implement, highly efficient, and very robust to differences in video content, allowing forinteractiveexploration of the temporal alignment space for multiple real world HD videos.

Given the growth of Internet photo collections, we now have a visual index of all major cities and tourist sites in the world. However, it is still a difficult task to capture that perfect shot with your own camera when visiting these places, especially when your camera itself has limitations, such as a limited field of view. In this paper, we propose a framework to overcome the imperfections of personal photographs of tourist sites using the rich information provided by large-scale Internet photo collections. Our method deploys state-of-the-art techniques for constructing initial 3D models from photo collections. The same techniques are then used to register personal photographs to these models, allowing us to augment personal 2D images with 3D information. This strong available scene prior allows us to address a number of traditionally challenging image enhancement techniques and achieve high-quality results using simple and robust algorithms. Specifically, we demonstrate automatic foreground segmentation, mono-to-stereo conversion, field-of-view expansion, photometric enhancement, and additionally automatic annotation with geolocation and tags. Our method clearly demonstrates some possible benefits of employing the rich information contained in online photo databases to efficiently enhance and augment one's own personal photographs.

Current trends in video technology indicate a significant increase in spatial and temporal resolution of video data. Recently, a linear-runtime feature diffusion algorithm was presented which aims for fast and accurate processing of such high resolution data. In this paper, we introduce this algorithm from the perspective of image-based depth estimation, expanding upon the algorithm by requiring interview consistency in the depth diffusion process. We also discuss different application scenarios and provide an in-depth analysis of the method in this context.

Live sports broadcast is seeing a large increase in the number of cameras used for filming. More cameras can provide better coverage of the field and a wider range of experiences for viewers. However, choosing optimal cameras for broadcast demands a high level of concentration, awareness and experience from sports broadcast directors. We present an automatic assistant to help select likely candidates from a large array of possible cameras. Sports directors can then choose the final broadcast camera from the reduced suggestion set. Our assistant uses both widely acknowledged cinematography guidelines for sports directing, as well as a data-driven approach that learns specific styles from directors.

Abstract We introduce a single unifying framework for a wide range of content‐aware image warping tasks using a finite element method (FEM). Existing approaches commonly define error terms over vertex finite differences and can be expressed as a special case of our general FEM model. In this work, we exploit the full generality of FEMs, gaining important advantages over prior methods. These advantages include arbitrary mesh connectivity allowing for adaptive meshing and efficient large‐scale solutions, a well‐defined continuous problem formulation that enables clear analysis of existing warping error functions and allows us to propose improved ones, and higher order basis functions that allow for smoother warps with fewer degrees of freedom. To support per‐element basis functions of varying degree and complex mesh connectivity with hanging nodes, we also introduce a novel use of discontinuous Galerkin FEM. We demonstrate the utility of our method by showing examples in video retargeting and camera stabilization applications, and compare our results with previous state of the art methods.

Abstract DuctTake is a system designed to enable practical compositing of multiple takes of a scene into a single video. Current industry solutions are based around object segmentation, a hard problem that requires extensive manual input and cleanup, making compositing an expensive part of the film‐making process. Our method instead composites shots together by finding optimal spatiotemporal seams using motion‐compensated 3D graph cuts through the video volume. We describe in detail the required components, decisions, and new techniques that together make a usable, interactive tool for compositing HD video, paying special attention to running time and performance of each section. We validate our approach by presenting a wide variety of examples and by comparing result quality and creation time to composites made by professional artists using current state‐of‐the‐art tools.

Today, stereoscopic 3D (S3D) cinema is already mainstream, and almost all new display devices for the home support S3D content. S3D distribution infrastructure to the home is already established partly in the form of 3D Blu-ray discs, video on demand services, or television channels. The necessity to wear glasses is, however, often considered as an obstacle, which hinders broader acceptance of this technology in the home. Multiviewautostereoscopic displays enable a glasses free perception of S3D content for several observers simultaneously, and support head motion parallax in a limited range. To support multiviewautostereoscopic displays in an already established S3D distribution infrastructure, a synthesis of new views from S3D video is needed. In this paper, a view synthesis method based on image-domain-warping (IDW) is presented that automatically synthesizes new views directly from S3D video and functions completely. IDW relies on an automatic and robust estimation of sparse disparities and image saliency information, and enforces target disparities in synthesized images using an image warping framework. Two configurations of the view synthesizer in the scope of a transmission and view synthesis framework are analyzed and evaluated. A transmission and view synthesis system that uses IDW is recently submitted to MPEG's call for proposals on 3D video technology, where it is ranked among the four best performing proposals.

One of the most fundamental problems in image processing and computer vision is the inherent ambiguity that exists between texture edges and object boundaries in real-world images and video. Despite this ambiguity, many applications in computer vision and image processing often use image edge strength with the assumption that these edges approximate object depth boundaries. However, this assumption is often invalidated by real world data, and this discrepancy is a significant limitation in many of today's image processing methods. We address this issue by introducing a simple, low-level, and patch-consistency assumption that leverages the extra information present in video data to resolve this ambiguity. Through analyzing how well patches can be modeled by simple transformations over time, we can obtain an indication of which image edges correspond to texture edges versus object boundaries. Our approach is simple to implement and has the potential to improve a wide range of image and video-based applications by suppressing the detrimental effects of strong texture edges on regularization terms. We validate our approach by presenting results on a variety of scene types and directly incorporating our augmented edge map into existing image segmentation and optical flow applications, showing results that better correspond to object boundaries.

We present an efficient and simple method for introducing temporal consistency to a large class of optimization driven image-based computer graphics problems. Our method extends recent work in edge-aware filtering, approximating costly global regularization with a fast iterative joint filtering operation. Using this representation, we can achieve tremendous efficiency gains both in terms of memory requirements and running time. This enables us to process entire shots at once, taking advantage of supporting information that exists across far away frames, something that is difficult with existing approaches due to the computational burden of video data. Our method is able to filter along motion paths using an iterative approach that simultaneously uses and estimates per-pixel optical flow vectors. We demonstrate its utility by creating temporally consistent results for a number of applications including optical flow, disparity estimation, colorization, scribble propagation, sparse data up-sampling, and visual saliency computation.

Given video from a single camera, conversion to two-view stereoscopic 3D is a challenging problem. We present a system to automatically create high quality stereoscopic video from monoscopic footage of field-based sports by exploiting context-specific priors, such as the ground plane, player size and known background. Our main contribution is a novel technique that constructs per-shot panoramas to ensure temporally consistent stereoscopic depth in video reconstructions. Players are rendered as billboards at correct depths on the ground plane. Our method uses additional sports priors to disambiguate segmentation artifacts and produce synthesized 3D shots that are in most cases, indistinguishable from stereoscopic ground truth footage.

Content creation for autostereoscopic displays is a widely unresolved task. Typical methods rely on view synthesis based on depth image based rendering. Our method applies purely image domain warping instead. Input video is analyzed and information about sparse disparity, vertical edges and saliency is extracted. A constrained energy minimization problem is formulated and efficiently solved. The resulting image warping functions are used to synthesize novel views. Our approach is fully automatic, accurate, and reliable. Disocclusions and related artifacts are avoided due to smooth, saliency-driven warping functions. Our method also works well for extrapolation of views in a limited range, thus supporting multiview creation from stereo input, which is the most relevant use case scenario.

This paper addresses the problem of remapping the disparity range of stereoscopic images and video. Such operations are highly important for a variety of issues arising from the production, live broadcast, and consumption of 3D content. Our work is motivated by the observation that the displayed depth and the resulting 3D viewing experience are dictated by a complex combination of perceptual, technological, and artistic constraints. We first discuss the most important perceptual aspects of stereo vision and their implications for stereoscopic content creation. We then formalize these insights into a set of basic disparity mapping operators. These operators enable us to control and retarget the depth of a stereoscopic scene in a nonlinear and locally adaptive fashion. To implement our operators, we propose a new strategy based on stereoscopic warping of the input video streams. From a sparse set of stereo correspondences, our algorithm computes disparity and image-based saliency estimates, and uses them to compute a deformation of the input views so as to meet the target disparities. Our approach represents a practical solution for actual stereo production and display that does not require camera calibration, accurate dense depth maps, occlusion handling, or inpainting. We demonstrate the performance and versatility of our method using examples from live action post-production, 3D display size adaptation, and live broadcast. An additional user study and ground truth comparison further provide evidence for the quality and practical relevance of the presented work.

Segmenting images into distinct material types is a very useful capability. Most work in image segmentation addresses the case where only a single image is available. Some methods improve on this by collecting HDR or multispectral images. However, it is also possible to use the reflectance properties of the materials to obtain better results. By acquiring many images of an object under different lighting conditions we have more samples of the surfaces bidirectional reflectance distribution function (BRDF). We show that this additional information enlarges the class of material types that can be well separated by segmentation, and that properly treating the information as samples of the BRDF further increases accuracy without requiring an explicit estimation of the material BRDF.

Abstract Inappropriate lighting is often responsible for poor quality video. In most offices and homes, lighting is not designed for video conferencing. This can result in unevenly lit faces, distracting shadows, and unnatural colors. We present a method for relighting faces that reduces the effects of uneven lighting and color. Our setup consists of a compact lighting rig and a camera that is both inexpensive and inconspicuous to the user. We use unperceivable infrared (IR) lights to obtain an illumination bases of the scene. Our algorithm computes an optimally weighted combination of IR bases to minimize lighting inconsistencies in foreground areas and reduce the effects of colored monitor light. However, IR relighting alone results in images with an unnatural ghostly appearance, thus a retargeting technique is presented which removes the unnatural IR effects and produces videos that have substantially more balanced intensity and color than the original video.

Video matting is the process of taking a sequence of frames, isolating the foreground, and replacing the background in each frame. We look at existing single-frame matting techniques and present a method that improves upon them by adding depth information acquired by a time-offlight range scanner. We use the depth information to automate the process so it can be practically used for video sequences. In addition, we show that we can improve the results from natural matting algorithms by adding a depth channel. The additional depth information allows us to reduce the artifacts that arise from ambiguities that occur when an object is a similar color to its background.

Abstract Hand held long exposures often result in blurred photographs due to camera shake. Long exposures are desirable both for artistic effect and in low‐light situations. We propose a novel method for digitally reducing imaging artifacts, which does not require additional hardware such as tripods or optical image stabilization lenses. A series of photographs is acquired with short shutter times, stabilized using image alignment, and then composited. Our method is capable of reducing noise and blurring due to camera shake, while simultaneously preserving the desirable effects of motion blur. The resulting images are very similar to those obtained using a tripod and a true extended exposure.