We propose a method for generating fly-through videos of a scene, from a single image and a given camera trajectory. We build upon an image-to-video latent diffusion model [5]. We condition its UNet [25] denoiser on the camera trajectory, using four techniques. (1) We condition UNet's temporal blocks on raw camera extrinsics, similar to MotionCtrl [36]. (2) We use images containing camera ray parameters, similar to CameraCtrl [14]. (3) We re-project the initial image to subsequent frames and condition on the resulting video. (4) We introduce a global 3D representation using 2D ⇔ 3D transformers [32], which implicitly conditions on the camera poses. We combine all conditions in a ContolNet-style [42] architecture. We then propose a metric that evaluates overall video quality and the ability to preserve details with view changes, which we use to analyze the trade-offs of individual and combined conditions. Finally, we identify an optimal combination of conditions. We calibrate camera positions in our datasets for scale consistency across scenes, and we train our scene exploration model, CamCtrl3D, demonstrating state-of-the-art results.

Despite the recent progress, existing frame interpolation methods still struggle with processing extremely high resolution input and handling challenging cases such as repetitive textures, thin objects, and large motion. To address these issues, we introduce a patch-based cascaded pixel diffusion model for high resolution frame interpolation, HiFI, that excels in these scenarios while achieving competitive performance on standard benchmarks. Cascades, which generate a series of images from low to high resolution, can help significantly with large or complex motion that require both global context for a coarse solution and detailed context for high resolution output. However, contrary to prior work on cascaded diffusion models which perform diffusion on increasingly large resolutions, we use a single model that always performs diffusion at the same resolution and upsamples by processing patches of the inputs and the prior solution. At inference time, this drastically reduces memory usage and allows a single model, solving both frame interpolation (base model’s task) and spatial up-sampling, saving training cost as well. HiFI excels at high-resolution images and complex repeated textures that require global context, achieving comparable or state-of-the-art performance on various benchmarks (Vimeo, Xiph, X-Test, and SEPE-8K). We further introduce a new dataset, LaMoR, that focuses on particularly challenging cases, and HiFI significantly outperforms other baselines.

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.

We present Magic Insert, a method for dragging-and-dropping subjects from a user-provided image into a target image of a different style in a physically plausible manner while matching the style of the target image. This work formalizes the problem of style-aware drag-and-drop and presents a method for tackling it by addressing two sub-problems: style-aware personalization and realistic object insertion in stylized images. For style-aware personalization, our method first fine-tunes a pretrained text-to-image diffusion model using LoRA and learned text tokens on the subject image, and then infuses it with a CLIP representation of the target style. For object insertion, we use Bootstrapped Domain Adaption to adapt a domain-specific photorealistic object insertion model to the domain of diverse artistic styles. Overall, the method significantly outperforms traditional approaches such as inpainting. Finally, we present a dataset, SubjectPlop, to facilitate evaluation and future progress in this area. Project page: https://magicinsert.github.io/

Scaling up autoregressive models in vision has not proven as beneficial as in large language models. In this work, we investigate this scaling problem in the context of text-to-image generation, focusing on two critical factors: whether models use discrete or continuous tokens, and whether tokens are generated in a random or fixed raster order using BERT- or GPT-like transformer architectures. Our empirical results show that, while all models scale effectively in terms of validation loss, their evaluation performance -- measured by FID, GenEval score, and visual quality -- follows different trends. Models based on continuous tokens achieves significantly better visual quality than those using discrete tokens. Furthermore, the generation order and attention mechanisms significantly affect the GenEval score: random-order models achieve notably better GenEval scores compared to raster-order models. Inspired by these findings, we train Fluid, a random-order autoregressive model on continuous tokens. Fluid 10.5B model achieves a new state-of-the-art zeor-shot FID of 6.16 on MS-COCO 30K, and 0.69 overall score on the GenEval benchmark. We hope our findings and results will encourage future efforts to further bridge the scaling gap between vision and language models.

We introduce the concept of a generative infinite game, a video game that transcends the traditional boundaries of finite, hard-coded systems by using generative models. Inspired by James P. Carse's distinction between finite and infinite games, we leverage recent advances in generative AI to create Unbounded: a game of character life simulation that is fully encapsulated in generative models. Specifically, Unbounded draws inspiration from sandbox life simulations and allows you to interact with your autonomous virtual character in a virtual world by feeding, playing with and guiding it - with open-ended mechanics generated by an LLM, some of which can be emergent. In order to develop Unbounded, we propose technical innovations in both the LLM and visual generation domains. Specifically, we present: (1) a specialized, distilled large language model (LLM) that dynamically generates game mechanics, narratives, and character interactions in real-time, and (2) a new dynamic regional image prompt Adapter (IP-Adapter) for vision models that ensures consistent yet flexible visual generation of a character across multiple environments. We evaluate our system through both qualitative and quantitative analysis, showing significant improvements in character life simulation, user instruction following, narrative coherence, and visual consistency for both characters and the environments compared to traditional related approaches.

Recent advances in large-scale pretraining have yielded visual foundation models with strong capabilities. Not only can recent models generalize to arbitrary images for their training task, their intermediate representations are useful for other visual tasks such as detection and segmentation. Given that such models can classify, delineate, and local- ize objects in 2D, we ask whether they also represent their 3D structure? In this work, we analyze the 3D awareness of visual foundation models. We posit that 3D awareness implies that representations (1) encode the 3D structure of the scene and (2) consistently represent the surface across views. We conduct a series of experiments using task-specific probes and zero-shot inference procedures on frozen fea- tures. Our experiments reveal several limitations of the current models. Our code and analysis can be found at https://github.com/mbanani/probe3d.

Personalization has emerged as a prominent aspect within the field of generative AI, enabling the synthesis of individuals in diverse contexts and styles, while retaining high-fidelity to their identities. However, the process of personalization presents inherent challenges in terms of time and memory requirements. Fine-tuning each personalized model needs considerable GPU time investment, and storing a personalized model per subject can be demanding in terms of storage capacity. To overcome these challenges, we propose HyperDreamBooth—a hypernetwork capable of efficiently generating a small set of personalized weights from a single image of a person. By composing these weights into the diffusion model, coupled with fast finetuning, HyperDreamBooth can generate a person's face in various contexts and styles, with high subject details while also preserving the model's crucial knowledge of diverse styles and semantic modifications. Our method achieves personalization on faces in roughly 20 seconds, 25x faster than DreamBooth and 125x faster than Textual Inversion, using as few as one reference image, with the same quality and style diversity as DreamBooth. Also our method yields a model that is 10,000x smaller than a normal DreamBooth model.

We introduce WonderJourney, a modular framework for perpetual 3D scene generation. Unlike prior work on view generation that focuses on a single type of scenes, we start at any user-provided location (by a text description or an image), and generate a journey through a long sequence of diverse yet coherently connected 3D scenes. We leverage an LLM to generate textual descriptions of the scenes in this journey, a text-driven point cloud generation pipeline to make a compelling and coherent sequence of 3D scenes, and a large VLM to verify the generated scenes. We show compelling, diverse visual results across various scene types and styles, forming imaginary “wonder journeys ”. Project website: https://kovenyu.com/WonderJourney/. “No, no! The adventures first, explanations take such a dreadful time.“ - Alice's Adventures in Wonderland

We propose a new approach for generative modeling based on training a neural network to be idempotent. An idempotent operator is one that can be applied sequentially without changing the result beyond the initial application, namely $f(f(z))=f(z)$. The proposed model $f$ is trained to map a source distribution (e.g, Gaussian noise) to a target distribution (e.g. realistic images) using the following objectives: (1) Instances from the target distribution should map to themselves, namely $f(x)=x$. We define the target manifold as the set of all instances that $f$ maps to themselves. (2) Instances that form the source distribution should map onto the defined target manifold. This is achieved by optimizing the idempotence term, $f(f(z))=f(z)$ which encourages the range of $f(z)$ to be on the target manifold. Under ideal assumptions such a process provably converges to the target distribution. This strategy results in a model capable of generating an output in one step, maintaining a consistent latent space, while also allowing sequential applications for refinement. Additionally, we find that by processing inputs from both target and source distributions, the model adeptly projects corrupted or modified data back to the target manifold. This work is a first step towards a ``global projector'' that enables projecting any input into a target data distribution.

Customizing text-to-image (T2I) models has seen tremendous progress recently, particularly in areas such as personalization, stylization, and conditional generation. However, expanding this progress to video generation is still in its infancy, primarily due to the lack of customized video data. In this work, we introduce Still-Moving, a novel generic framework for customizing a text-to-video (T2V) model, without requiring any customized video data. The framework applies to the prominent T2V design where the video model is built over a T2I model (e.g., via inflation). We assume access to a customized version of the T2I model, trained only on still image data (e.g., using DreamBooth). Naively plugging in the weights of the customized T2I model into the T2V model often leads to significant artifacts or insufficient adherence to the customization data. To overcome this issue, we train lightweight Spatial Adapters that adjust the features produced by the injected T2I layers. Importantly, our adapters are trained on "frozen videos" (i.e., repeated images), constructed from image samples generated by the customized T2I model. This training is facilitated by a novel Motion Adapter module, which allows us to train on such static videos while preserving the motion prior of the video model. At test time, we remove the Motion Adapter modules and leave in only the trained Spatial Adapters. This restores the motion prior of the T2V model while adhering to the spatial prior of the customized T2I model. We demonstrate the effectiveness of our approach on diverse tasks including personalized, stylized, and conditional generation. In all evaluated scenarios, our method seamlessly integrates the spatial prior of the customized T2I model with a motion prior supplied by the T2V model.

Recent advances in generative imagery have brought forth outpainting and inpainting models that can produce high-quality, plausible image content in unknown regions. However, the content these models hallucinate is necessarily inauthentic, since they are unaware of the true scene. In this work, we propose RealFill, a novel generative approach for image completion that fills in missing regions of an image with the content that should have been there. RealFill is a generative inpainting model that is personalized using only a few reference images of a scene. These reference images do not have to be aligned with the target image, and can be taken with drastically varying viewpoints, lighting conditions, camera apertures, or image styles. Once personalized, RealFill is able to complete a target image with visually compelling contents that are faithful to the original scene. We evaluate RealFill on a new image completion benchmark that covers a set of diverse and challenging scenarios, and find that it outperforms existing approaches by a large margin. Project page: https://realfill.github.io.

Scene flow estimation is a long-standing problem in computer vision, where the goal is to find the 3D motion of a scene from its consecutive observations. Recently, there have been efforts to compute the scene flow from 3D point clouds. A common approach is to train a regression model that consumes source and target point clouds and outputs the per-point translation vector. An alternative is to learn point matches between the point clouds concurrently with regressing a refinement of the initial correspondence flow. In both cases, the learning task is very challenging since the flow regression is done in the free 3D space, and a typical solution is to resort to a large annotated synthetic dataset. We introduce SCOOP, a new method for scene flow estimation that can be learned on a small amount of data without employing ground-truth flow supervision. In contrast to previous work, we train a pure correspondence model focused on learning point feature representation and initialize the flow as the difference between a source point and its softly corresponding target point. Then, in the run-time phase, we directly optimize a flow refinement component with a self-supervised objective, which leads to a coherent and accurate flow field between the point clouds. Experiments on widespread datasets demonstrate the performance gains achieved by our method compared to existing leading techniques while using a fraction of the training data. Our code is publicly available11https://github.com/itailang/SCOOP .

Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for “personalization” of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/

Automatically estimating 3D skeleton, shape, camera viewpoints, and part articulation from sparse in-the-wild image ensembles is a severely under-constrained and challenging problem. Most prior methods rely on large-scale image datasets, dense temporal correspondence, or human annotations like camera pose, 2D keypoints, and shape templates. We propose Hi-LASSIE, which performs 3D articulated reconstruction from only 20–30 online images in the wild without any user-defined shape or skeleton templates. We follow the recent work of LASSIE that tackles a similar problem setting and make two significant advances. First, instead of relying on a manually annotated 3D skeleton, we automatically estimate a class-specific skeleton from the selected reference image. Second, we improve the shape reconstructions with novel instance-specific optimization strategies that allow reconstructions to faithful fit on each instance while preserving the class-specific priors learned across all images. Experiments on in-the-wild image ensembles show that Hi-LASSIE obtains higher fidelity state-of-the-art 3D reconstructions despite requiring minimum user input. Project page: chhankyao.github.io/hi-lassie/

Motion magnification helps us visualize subtle, imperceptible motion. However, prior methods only work for 2D videos captured with a fixed camera. We present a 3D motion magnification method that can magnify subtle motions from scenes captured by a moving camera, while supporting novel view rendering. We represent the scene with time-varying radiance fields and leverage the Eulerian principle for motion magnification to extract and amplify the variation of the embedding of a fixed point over time. We study and validate our proposed principle for 3D motion magnification using both implicit and tri-plane-based radiance fields as our underlying 3D scene representation. We evaluate the effectiveness of our method on both synthetic and real-world scenes captured under various camera setups.

Priors are essential for reconstructing images from noisy and/or incomplete measurements. The choice of the prior determines both the quality and uncertainty of recovered images. We propose turning score-based diffusion models into principled image priors ("score-based priors") for analyzing a posterior of images given measurements. Previously, probabilistic priors were limited to handcrafted regularizers and simple distributions. In this work, we empirically validate the theoretically-proven probability function of a score-based diffusion model. We show how to sample from resulting posteriors by using this probability function for variational inference. Our results, including experiments on denoising, deblurring, and interferometric imaging, suggest that score-based priors enable principled inference with a sophisticated, data-driven image prior.

We present DreamBooth3D, an approach to personalize text-to-3D generative models from as few as 3-6 casually captured images of a subject. Our approach combines recent advances in personalizing text-to-image models (DreamBooth) with text-to-3D generation (DreamFusion). We find that naïvely combining these methods fails to yield satisfactory subject-specific 3D assets due to personalized text-to-image models overfitting to the input viewpoints of the subject. We overcome this through a 3-stage optimization strategy where we jointly leverage the 3D consistency of neural radiance fields together with the personalization capability of text-to-image models. Our method can produce high-quality, subject-specific 3D assets with text-driven modifications such as novel poses, colors and attributes that are not seen in any of the input images of the subject. More results are available at our project page: https://dreambooth3d.github.io

We present Muse, a text-to-image Transformermodel that achieves state-of-the-art image genera-tion performance while being significantly moreefficient than diffusion or autoregressive models.Muse is trained on a masked modeling task indiscrete token space: given the text embeddingextracted from a pre-trained large language model(LLM), Muse learns to predict randomly maskedimage tokens. Compared to pixel-space diffusionmodels, such as Imagen and DALL-E 2, Muse issignificantly more efficient due to the use of dis-crete tokens and requires fewer sampling itera-tions; compared to autoregressive models such asParti, Muse is more efficient due to the use of par-allel decoding. The use of a pre-trained LLM en-ables fine-grained language understanding, whichtranslates to high-fidelity image generation andthe understanding of visual concepts such as ob-jects, their spatial relationships, pose, cardinalityetc. Our 900M parameter model achieves a newSOTA on CC3M, with an FID score of 6.06. TheMuse 3B parameter model achieves an FID of7.88 on zero-shot COCO evaluation, along with aCLIP score of 0.32. Muse also directly enables anumber of image editing applications without theneed to fine-tune or invert the model: inpainting,outpainting, and mask-free editing. More resultsand videos demonstrating editing are available at https://muse-icml.github.io/

Pre-trained large text-to-image models synthesize impressive images with an appropriate use of text prompts. However, ambiguities inherent in natural language, and out-of-distribution effects make it hard to synthesize arbitrary image styles, leveraging a specific design pattern, texture or material. In this paper, we introduce *StyleDrop*, a method that enables the synthesis of images that faithfully follow a specific style using a text-to-image model. StyleDrop is extremely versatile and captures nuances and details of a user-provided style, such as color schemes, shading, design patterns, and local and global effects. StyleDrop works by efficiently learning a new style by fine-tuning very few trainable parameters (less than 1\% of total model parameters), and improving the quality via iterative training with either human or automated feedback. Better yet, StyleDrop is able to deliver impressive results even when the user supplies only a *single* image specifying the desired style. An extensive study shows that, for the task of style tuning text-to-image models, StyleDrop on Muse convincingly outperforms other methods, including DreamBooth and textual inversion on Imagen or Stable Diffusion. More results are available at our project website: [https://styledrop.github.io](https://styledrop.github.io).

Estimating 3D articulated shapes like animal bodies from monocular images is inherently challenging due to the ambiguities of camera viewpoint, pose, texture, lighting, etc. We propose ARTIC3D, a self-supervised framework to reconstruct per-instance 3D shapes from a sparse image collection in-the-wild. Specifically, ARTIC3D is built upon a skeleton-based surface representation and is further guided by 2D diffusion priors from Stable Diffusion. First, we enhance the input images with occlusions/truncation via 2D diffusion to obtain cleaner mask estimates and semantic features. Second, we perform diffusion-guided 3D optimization to estimate shape and texture that are of high-fidelity and faithful to input images. We also propose a novel technique to calculate more stable image-level gradients via diffusion models compared to existing alternatives. Finally, we produce realistic animations by fine-tuning the rendered shape and texture under rigid part transformations. Extensive evaluations on multiple existing datasets as well as newly introduced noisy web image collections with occlusions and truncation demonstrate that ARTIC3D outputs are more robust to noisy images, higher quality in terms of shape and texture details, and more realistic when animated.

Using only a model that was trained to predict where people look at images, and no additional training data, we can produce a range of powerful editing effects for reducing distraction in images. Given an image and a mask specifying the region to edit, we backpropagate through a state-of-the-art saliency model to parameterize a differentiable editing operator, such that the saliency within the masked region is reduced. We demonstrate several operators, including: a recoloring operator, which learns to apply a color transform that camouflages and blends distractors into their surroundings; a warping operator, which warps less salient image regions to cover distractors, gradually collapsing objects into themselves and effectively removing them (an effect akin to inpainting); a GAN operator, which uses a semantic prior to fully replace image regions with plausible, less salient alternatives. The resulting effects are consistent with cognitive research on the human visual system (e.g., since color mismatch is salient, the recoloring operator learns to harmonize objects' colors with their surrounding to reduce their saliency). And importantly, all effects are achieved under a zero-shot learning scenario, solely through the guidance of the pretrained saliency model, with no supervised data of the effects. We present results on a variety of natural images and conduct a perceptual study to evaluate and validate the changes in viewers' eye-gaze between the original images and our edited results. Project Webpage: https://deep-saliency-prior.github.io/

We explore a feed-forward approach for decomposing a video into layers, where each layer contains an object of interest along with its associated shadows, reflections, and other visual effects. This problem is challenging since associated effects vary widely with the 3D geometry and lighting conditions in the scene, and ground-truth labels for visual effects are difficult (and in some cases impractical) to collect. We take a self-supervised approach and train a neural network to produce a foreground image and alpha matte from a rough object segmentation mask under a reconstruction and sparsity loss. Under reconstruction loss, the layer decomposition problem is underdetermined: many combinations of layers may reconstruct the input video.Inspired by the game theory concept of focal points---or \emph{Schelling points}---we pose the problem as a coordination game, where each player (network) predicts the effects for a single object without knowledge of the other players' choices. The players learn to converge on the ``natural'' layer decomposition in order to maximize the likelihood of their choices aligning with the other players'. We train the network to play this game with itself, and show how to design the rules of this game so that the focal point lies at the correct layer decomposition. We demonstrate feed-forward results on a challenging synthetic dataset, then show that pretraining on this dataset significantly reduces optimization time for real videos.

Creating high-quality articulated 3D models of animals is challenging either via manual creation or using 3D scanning tools. Therefore, techniques to reconstruct articulated 3D objects from 2D images are crucial and highly useful. In this work, we propose a practical problem setting to estimate 3D pose and shape of animals given only a few (10-30) in-the-wild images of a particular animal species (say, horse). Contrary to existing works that rely on pre-defined template shapes, we do not assume any form of 2D or 3D ground-truth annotations, nor do we leverage any multi-view or temporal information. Moreover, each input image ensemble can contain animal instances with varying poses, backgrounds, illuminations, and textures. Our key insight is that 3D parts have much simpler shape compared to the overall animal and that they are robust w.r.t. animal pose articulations. Following these insights, we propose LASSIE, a novel optimization framework which discovers 3D parts in a self-supervised manner with minimal user intervention. A key driving force behind LASSIE is the enforcing of 2D-3D part consistency using self-supervisory deep features. Experiments on Pascal-Part and self-collected in-the-wild animal datasets demonstrate considerably better 3D reconstructions as well as both 2D and 3D part discovery compared to prior arts. Project page: https://chhankyao.github.io/lassie/

Computer vision is increasingly effective at segmenting objects in images and videos; however, scene effects related to the objects—shadows, reflections, generated smoke, etc.—are typically overlooked. Identifying such scene effects and associating them with the objects producing them is important for improving our fundamental understanding of visual scenes, and can also assist a variety of applications such as removing, duplicating, or enhancing objects in video. In this work, we take a step towards solving this novel problem of automatically associating objects with their effects in video. Given an ordinary video and a rough segmentation mask over time of one or more subjects of interest, we estimate an omnimatte for each subject—an alpha matte and color image that includes the subject along with all its related time-varying scene elements. Our model is trained only on the input video in a self-supervised manner, without any manual labels, and is generic—it produces omnimattes automatically for arbitrary objects and a variety of effects. We show results on real-world videos containing interactions between different types of subjects (cars, animals, people) and complex effects, ranging from semitransparent elements such as smoke and reflections, to fully opaque effects such as objects attached to the subject.1

We wish to automatically predict the “speediness” of moving objects in videos - whether they move faster, at, or slower than their “natural” speed. The core component in our approach is SpeedNet - a novel deep network trained to detect if a video is playing at normal rate, or if it is sped up. SpeedNet is trained on a large corpus of natural videos in a self-supervised manner, without requiring any manual annotations. We show how this single, binary classification network can be used to detect arbitrary rates of speediness of objects. We demonstrate prediction results by SpeedNet on a wide range of videos containing complex natural motions, and examine the visual cues it utilizes for making those predictions. Importantly, we show that through predicting the speed of videos, the model learns a powerful and meaningful space-time representation that goes beyond simple motion cues. We demonstrate how those learned features can boost the performance of self supervised action recognition, and can be used for video retrieval. Furthermore, we also apply SpeedNet for generating time-varying, adaptive video speedups, which can allow viewers to watch videos faster, but with less of the jittery, unnatural motions typical to videos that are sped up uniformly.

We present a method for retiming people in an ordinary, natural video --- manipulating and editing the time in which different motions of individuals in the video occur. We can temporally align different motions, change the speed of certain actions (speeding up/slowing down, or entirely "freezing" people), or "erase" selected people from the video altogether. We achieve these effects computationally via a dedicated learning-based layered video representation, where each frame in the video is decomposed into separate RGBA layers, representing the appearance of different people in the video. A key property of our model is that it not only disentangles the direct motions of each person in the input video, but also correlates each person automatically with the scene changes they generate---e.g., shadows, reflections, and motion of loose clothing. The layers can be individually retimed and recombined into a new video, allowing us to achieve realistic, high-quality renderings of retiming effects for real-world videos depicting complex actions and involving multiple individuals, including dancing, trampoline jumping, or group running.

How much can we infer about a person’s looks from the way they speak? In this paper, we study the task of reconstructing a facial image of a person from a short audio recording of that person speaking. We design and train a deep neural network to perform this task using millions of natural Internet/Youtube videos of people speaking. During training, our model learns voice-face correlations that allow it to produce images that capture various physical attributes of the speakers such as age, gender and ethnicity. This is done in a self-supervised manner, by utilizing the natural co-occurrence of faces and speech in Internet videos, without the need to model attributes explicitly. We evaluate and numerically quantify how–-and in what manner–-our Speech2Face reconstructions, obtained directly from audio, resemble the true face images of the speakers.

We present a joint audio-visual model for isolating a single speech signal from a mixture of sounds such as other speakers and background noise. Solving this task using only audio as input is extremely challenging and does not provide an association of the separated speech signals with speakers in the video. In this paper, we present a deep network-based model that incorporates both visual and auditory signals to solve this task. The visual features are used to "focus" the audio on desired speakers in a scene and to improve the speech separation quality. To train our joint audio-visual model, we introduce AVS peech , a new dataset comprised of thousands of hours of video segments from the Web. We demonstrate the applicability of our method to classic speech separation tasks, as well as real-world scenarios involving heated interviews, noisy bars, and screaming children, only requiring the user to specify the face of the person in the video whose speech they want to isolate. Our method shows clear advantage over state-of-the-art audio-only speech separation in cases of mixed speech. In addition, our model, which is speaker-independent (trained once, applicable to any speaker), produces better results than recent audio-visual speech separation methods that are speaker-dependent (require training a separate model for each speaker of interest).

Visible watermarking is a widely-used technique for marking and protecting copyrights of many millions of images on the web, yet it suffers from an inherent security flaw-watermarks are typically added in a consistent manner to many images. We show that this consistency allows to automatically estimate the watermark and recover the original images with high accuracy. Specifically, we present a generalized multi-image matting algorithm that takes a watermarked image collection as input and automatically estimates the foreground (watermark), its alpha matte, and the background (original) images. Since such an attack relies on the consistency of watermarks across image collection, we explore and evaluate how it is affected by various types of inconsistencies in the watermark embedding that could potentially be used to make watermarking more secured. We demonstrate the algorithm on stock imagery available on the web, and provide extensive quantitative analysis on synthetic watermarked data. A key takeaway message of this paper is that visible watermarks should be designed to not only be robust against removal from a single image, but to be more resistant to mass-scale removal from image collections as well.

The estimation of material properties is important for scene understanding, with many applications in vision, robotics, and structural engineering. This paper connects fundamentals of vibration mechanics with computer vision techniques in order to infer material properties from small, often imperceptible motions in video. Objects tend to vibrate in a set of preferred modes. The frequencies of these modes depend on the structure and material properties of an object. We show that by extracting these frequencies from video of a vibrating object, we can often make inferences about that object's material properties. We demonstrate our approach by estimating material properties for a variety of objects by observing their motion in high-speed and regular frame rate video.

We present a principled framework for inferring pixel labels in weakly-annotated image datasets. Most previous, example-based approaches to computer vision rely on a large corpus of densely labeled images. However, for large, modern image datasets, such labels are expensive to obtain and are often unavailable. We establish a large-scale graphical model spanning all labeled and unlabeled images, then solve it to infer pixel labels jointly for all images in the dataset while enforcing consistent annotations over similar visual patterns. This model requires significantly less labeled data and assists in resolving ambiguities by propagating inferred annotations from images with stronger local visual evidences to images with weaker local evidences. We apply our proposed framework to two computer vision problems, namely image annotation with semantic segmentation, and object discovery and co-segmentation (segmenting multiple images containing a common object). Extensive numerical evaluations and comparisons show that our method consistently outperforms the state-of-the-art in automatic annotation and semantic labeling, while requiring significantly less labeled data. In contrast to previous co-segmentation techniques, our method manages to discover and segment objects well even in the presence of substantial amounts of noise images (images not containing the common object), as typical for datasets collected from Internet search.

The estimation of material properties is important for scene understanding, with many applications in vision, robotics, and structural engineering. This paper connects fundamentals of vibration mechanics with computer vision techniques in order to infer material properties from small, often imperceptible motion in video. Objects tend to vibrate in a set of preferred modes. The shapes and frequencies of these modes depend on the structure and material properties of an object. Focusing on the case where geometry is known or fixed, we show how information about an object's modes of vibration can be extracted from video and used to make inferences about that object's material properties. We demonstrate our approach by estimating material properties for a variety of rods and fabrics by passively observing their motion in high-speed and regular framerate video.

We propose a novel method for template matching in unconstrained environments. Its essence is the Best-Buddies Similarity (BBS), a useful, robust, and parameter-free similarity measure between two sets of points. BBS is based on counting the number of Best-Buddies Pairs (BBPs)-pairs of points in source and target sets, where each point is the nearest neighbor of the other. BBS has several key features that make it robust against complex geometric deformations and high levels of outliers, such as those arising from background clutter and occlusions. We study these properties, provide a statistical analysis that justifies them, and demonstrate the consistent success of BBS on a challenging real-world dataset.

We present a unified computational approach for taking photos through reflecting or occluding elements such as windows and fences. Rather than capturing a single image, we instruct the user to take a short image sequence while slightly moving the camera. Differences that often exist in the relative position of the background and the obstructing elements from the camera allow us to separate them based on their motions, and to recover the desired background scene as if the visual obstructions were not there. We show results on controlled experiments and many real and practical scenarios, including shooting through reflections, fences, and raindrop-covered windows.

We present a new compact image pyramid representation, the Riesz pyramid, that can be used for real-time phase-based motion magnification. Our new representation is less overcomplete than even the smallest two orientation, octave-bandwidth complex steerable pyramid, and can be implemented using compact, efficient linear filters in the spatial domain. Motion-magnified videos produced with this new representation are of comparable quality to those produced with the complex steerable pyramid. When used with phase-based video magnification, the Riesz pyramid phase-shifts image features along only their dominant orientation rather than every orientation like the complex steerable pyramid.

When sound hits an object, it causes small vibrations of the object's surface. We show how, using only high-speed video of the object, we can extract those minute vibrations and partially recover the sound that produced them, allowing us to turn everyday objects---a glass of water, a potted plant, a box of tissues, or a bag of chips---into visual microphones. We recover sounds from high-speed footage of a variety of objects with different properties, and use both real and simulated data to examine some of the factors that affect our ability to visually recover sound. We evaluate the quality of recovered sounds using intelligibility and SNR metrics and provide input and recovered audio samples for direct comparison. We also explore how to leverage the rolling shutter in regular consumer cameras to recover audio from standard frame-rate videos, and use the spatial resolution of our method to visualize how sound-related vibrations vary over an object's surface, which we can use to recover the vibration modes of an object.

We present a new unsupervised algorithm to discover and segment out common objects from large and diverse image collections. In contrast to previous co-segmentation methods, our algorithm performs well even in the presence of significant amounts of noise images (images not containing a common object), as typical for datasets collected from Internet search. The key insight to our algorithm is that common object patterns should be salient within each image, while being sparse with respect to smooth transformations across other images. We propose to use dense correspondences between images to capture the sparsity and visual variability of the common object over the entire database, which enables us to ignore noise objects that may be salient within their own images but do not commonly occur in others. We performed extensive numerical evaluation on established co-segmentation datasets, as well as several new datasets generated using Internet search. Our approach is able to effectively segment out the common object for diverse object categories, while naturally identifying images where the common object is not present.

We introduce a technique to manipulate small movements in videos based on an analysis of motion in complex-valued image pyramids. Phase variations of the coefficients of a complex-valued steerable pyramid over time correspond to motion, and can be temporally processed and amplified to reveal imperceptible motions, or attenuated to remove distracting changes. This processing does not involve the computation of optical flow, and in comparison to the previous Eulerian Video Magnification method it supports larger amplification factors and is significantly less sensitive to noise. These improved capabilities broaden the set of applications for motion processing in videos. We demonstrate the advantages of this approach on synthetic and natural video sequences, and explore applications in scientific analysis, visualization and video enhancement.

Although dense, long-range, motion trajectories are a prominent representation of motion in videos, there is still no good solution for constructing dense motion tracks in a truly long-range fashion. Ideally, we would want every scene feature that appears in multiple, not necessarily contiguous, parts of the sequence to be associated with the same motion track. Despite this reasonable and clearly stated objective, there has been surprisingly little work on general-purpose algorithms that can accomplish this task. State-of-the-art dense motion trackers process the sequence incrementally in a frame-by-frame manner, and associate, by design, features that disappear and reappear in the video, with different tracks, thereby losing important information of the long-term motion signal. In this paper, we strive towards an algorithm for producing generic long-range motion trajectories that are robust to occlusion, deformation and camera motion. We leverage accurate local (short-range) trajectories produced by current motion tracking methods and use them as an initial estimate for a global (long-range) solution. Our algorithm re-correlates the short trajectories and links them to form a long-range motion representation by formulating a combinatorial assignment problem that is defined and optimized globally over the entire sequence. This allows to correlate features in arbitrarily distinct parts of the sequence, as well as handle tracking ambiguities by spatiotemporal regularization. We report the results of the algorithm on both synthetic and natural videos, and evaluate the long-range motion representation for action recognition.

Our goal is to reveal temporal variations in videos that are difficult or impossible to see with the naked eye and display them in an indicative manner. Our method, which we call Eulerian Video Magnification, takes a standard video sequence as input, and applies spatial decomposition, followed by temporal filtering to the frames. The resulting signal is then amplified to reveal hidden information. Using our method, we are able to visualize the flow of blood as it fills the face and also to amplify and reveal small motions. Our technique can run in real time to show phenomena occurring at the temporal frequencies selected by the user.

Motions can occur over both short and long time scales. We introduce motion denoising, which treats short-term changes as noise, long-term changes as signal, and re-renders a video to reveal the underlying long-term events. We demonstrate motion denoising for time-lapse videos. One of the characteristics of traditional time-lapse imagery is stylized jerkiness, where short-term changes in the scene appear as small and annoying jitters in the video, often obfuscating the underlying temporal events of interest. We apply motion denoising for resynthesizing time-lapse videos showing the long-term evolution of a scene with jerky short-term changes removed. We show that existing filtering approaches are often incapable of achieving this task, and present a novel computational approach to denoise motion without explicit motion analysis. We demonstrate promising experimental results on a set of challenging time-lapse sequences.

The numerous works on media retargeting call for a methodological approach for evaluating retargeting results. We present the first comprehensive perceptual study and analysis of image retargeting. First, we create a benchmark of images and conduct a large scale user study to compare a representative number of state-of-the-art retargeting methods. Second, we present analysis of the users' responses, where we find that humans in general agree on the evaluation of the results and show that some retargeting methods are consistently more favorable than others. Third, we examine whether computational image distance metrics can predict human retargeting perception. We show that current measures used in this context are not necessarily consistent with human rankings, and demonstrate that better results can be achieved using image features that were not previously considered for this task. We also reveal specific qualities in retargeted media that are more important for viewers. The importance of our work lies in promoting better measures to assess and guide retargeting algorithms in the future. The full benchmark we collected, including all images, retargeted results, and the collected user data, are available to the research community for further investigation at http://people.csail.mit.edu/mrub/retargetme.

Content aware resizing gained popularity lately and users can now choose from a battery of methods to retarget their media. However, no single retargeting operator performs well on all images and all target sizes. In a user study we conducted, we found that users prefer to combine seam carving with cropping and scaling to produce results they are satisfied with. This inspires us to propose an algorithm that combines different operators in an optimal manner. We define a resizing space as a conceptual multi-dimensional space combining several resizing operators, and show how a path in this space defines a sequence of operations to retarget media. We define a new image similarity measure, which we term Bi-Directional Warping (BDW), and use it with a dynamic programming algorithm to find an optimal path in the resizing space. In addition, we show a simple and intuitive user interface allowing users to explore the resizing space of various image sizes interactively. Using key-frames and interpolation we also extend our technique to retarget video, providing the flexibility to use the best combination of operators at different times in the sequence.

Video, like images, should support content aware resizing. We present video retargeting using an improved seam carving operator. Instead of removing 1D seams from 2D images we remove 2D seam manifolds from 3D space-time volumes. To achieve this we replace the dynamic programming method of seam carving with graph cuts that are suitable for 3D volumes. In the new formulation, a seam is given by a minimal cut in the graph and we show how to construct a graph such that the resulting cut is a valid seam. That is, the cut is monotonic and connected. In addition, we present a novel energy criterion that improves the visual quality of the retargeted images and videos. The original seam carving operator is focused on removing seams with the least amount of energy, ignoring energy that is introduced into the images and video by applying the operator. To counter this, the new criterion is looking forward in time - removing seams that introduce the least amount of energy into the retargeted result. We show how to encode the improved criterion into graph cuts (for images and video) as well as dynamic programming (for images). We apply our technique to images and videos and present results of various applications.

We give a formula for the number of interior intersection points made by the diagonals of a regular n-gon. The answer is a polynomial on each residue class modulo 2520. We also compute the number of regions formed by the diagonals, by using Euler's formula V - E + F = 2.

Binary m-sequences are maximal-length sequences generated by shift registers of length m, that are employed in navigation, radar, and spread-spectrum communication. It is well known that given a pair of distinct m-sequences, the crosscorrelation function must take on at least three values. This correspondence addresses a conjecture made by Helleseth in 1976, that if m is a power of 2, then there are no pairs of binary m-sequences with a 3-valued crosscorrelation function. This conjecture is proved under the assumption that the three correlation values are symmetric about -1.