Unsupervised optical flow methods have become more popular in the last decade, enabling the training of models across domains without ground truth data. Although RAFT and its successors have achieved significant success in the supervised settings, many unsupervised approaches continue to use older backbones such as PWC-Net. One reason for this architectural stagnation is that the current RAFT-based SOTA approach has proven challenging for the community to reproduce. In this paper, we revive and advance unsupervised optical flow: First, we introduce Sun-RAFT: a simple unsupervised RAFT. Second, building on Sun-RAFT, we present Muun-RAFT: a novel multi-scale unsupervised RAFT, where we propose a gradual context-based upsampling to refine the flow, further improving both accuracy and preservation of details. Third, we reexamine previously advised unsupervised strategies to identify effective training settings. In terms of results, both our methods demonstrate strong generalization capabilities and set a new SOTA for unsupervised two-frame approaches on MPI-Sintel, with Muun-RAFT surpassing even the current multi-frame SOTA by up to 28%. Finally, we open-source our PyTorch code, enabling further developments in the field: https://cv-stuttgart.github.io/Reviving-Unsupervised-OpticalFlow.

Vision language models (VLMs) respond to user-crafted text prompts and visual inputs, and are applied to numerous real-world problems. VLMs integrate visual modalities with large language models (LLMs), which are well known to be prompt-sensitive. Hence, it is crucial to determine whether VLMs inherit this instability to varying prompts. We therefore investigate which prompt variations VLMs are most sensitive to and which VLMs are most agnostic to prompt variations. To this end, we introduce PARC (Prompt Analysis via Reliability and Calibration), a VLM prompt sensitivity analysis framework built on three pillars: (1) plausible prompt variations in both the language and vision domain, (2) a novel model reliability score with built-in guarantees, and (3) a calibration step that enables dataset-and prompt-spanning prompt variation analysis. Regarding prompt variations, PARC’s evaluation shows that VLMs mirror LLM language prompt sensitivity in the vision domain, and most destructive variations change the expected answer. Regarding models, outstandingly robust VLMs among 22 evaluated models come from the InternVL2 family. We further find indications that prompt sensitivity is linked to training data. https://github.com/NVlabs/PARC

Although multi-scale concepts have recently proven useful for recurrent network architectures in the field of optical flow and stereo, they have not been considered for image-based scene flow so far. Hence, based on a single-scale recurrent scene flow backbone, we develop a multi-scale approach that generalizes successful hierarchical ideas from optical flow to image-based scene flow. By considering suitable concepts for the feature and the context encoder, the overall coarse-to-fine framework and the training loss, we succeed to design a scene flow approach that outperforms the current state of the art on KITTI and Spring by 8.7% (3.89 vs. 4.26) and 65.8% (9.13 vs. 26.71), respectively. Our code is available at https://github.com/cv-stuttgart/MS-RAFT-3D.

Attention-based motion aggregation concepts have recently shown their usefulness in optical flow estimation, in particular when it comes to handling occluded regions. However, due to their complexity, such concepts have been mainly restricted to coarse-resolution single-scale approaches that fail to provide the detailed outcome of high-resolution multi-scale networks. In this paper, we hence propose CCMR: a high-resolution coarse-to-fine approach that leverages attention-based motion grouping concepts to multi-scale optical flow estimation. CCMR relies on a hierarchical two-step attention-based context-motion grouping strategy that first computes global multi-scale context features and then uses them to guide the actual motion grouping. As we iterate both steps over all coarse-to-fine scales, we adapt cross covariance image transformers to allow for an efficient realization while maintaining scale-dependent properties. Experiments and ablations demonstrate that our efforts of combining multi-scale and attention-based concepts pay off. By providing highly detailed flow fields with strong improvements in both occluded and non-occluded regions, our CCMR approach not only outperforms both the corresponding single-scale attention-based and multi-scale attention-free baselines by up to 23.0% and 21.6%, respectively, it also achieves state-of-the-art results, ranking first on KITTI 2015 and second on MPI Sintel Clean and Final. Code and trained models are available at https://github.com/cv-stuttgart/CCMR.

Despite of exciting advances in image-based rendering and novel view synthesis, it is still challenging to achieve high-resolution results that can reach production-level quality when applying such methods to the task of stereo conversion. At the same time, only very few dedicated stereo conversion approaches exist, which also fall short in terms of the required quality. Hence, in this paper, we present a novel method for high-resolution 2D-to-3D conversion. It is fully differentiable in all of its stages and performs disparity-informed warping, consistent foreground-background compositing, and background-aware inpainting. To enable temporal consistency in the resulting video, we propose a strategy to integrate information from additional video frames. Extensive ablation studies validate our design choices, leading to a fully automatic model that outperforms existing approaches by a large margin (49-70% LPIPS error reduction). Finally, inspired from current practices in manual stereo conversion, we introduce optional interactive tools into our model, which allow to steer the conversion process and make it significantly more applicable for 3D film production.

Adversarial patches undermine the reliability of optical flow predictions when placed in arbitrary scene locations. Therefore, they pose a realistic threat to real-world motion detection and its downstream applications. Potential remedies are defense strategies that detect and remove adversarial patches, but their influence on the underlying motion prediction has not been investigated. In this paper, we thoroughly examine the currently available detect-and-remove defenses ILP and LGS for a wide selection of state-of-the-art optical flow methods, and illuminate their side effects on the quality and robustness of the final flow predictions. In particular, we implement defense-aware attacks to investigate whether current defenses are able to withstand attacks that take the defense mechanism into account. Our experiments yield two surprising results: Detect-and-remove defenses do not only lower the optical flow quality on benign scenes, in doing so, they also harm the robustness under patch attacks for all tested optical flow methods except FlowNetC. As currently employed detect-and-remove defenses fail to deliver the promised adversarial robustness for optical flow, they evoke a false sense of security. The code is available at https://github.com/cvstuttgart/DetectionDefenses.

Abstract Hierarchical concepts have proven useful in many classical and learning-based optical flow methods regarding both accuracy and robustness. In this paper we show that such concepts are still useful in the context of recent neural networks that follow RAFT’s paradigm refraining from hierarchical strategies by relying on recurrent updates based on a single-scale all-pairs transform. To this end, we introduce MS-RAFT+: a novel recurrent multi-scale architecture based on RAFT that unifies several successful hierarchical concepts. It employs a coarse-to-fine estimation to enable the use of finer resolutions by useful initializations from coarser scales. Moreover, it relies on RAFT’s correlation pyramid that allows to consider non-local cost information during the matching process. Furthermore, it makes use of advanced multi-scale features that incorporate high-level information from coarser scales. And finally, our method is trained subject to a sample-wise robust multi-scale multi-iteration loss that closely supervises each iteration on each scale, while allowing to discard particularly difficult samples. In combination with an appropriate mixed-dataset training strategy, our method performs favorably. It not only yields highly accurate results on the four major benchmarks (KITTI 2015, MPI Sintel, Middlebury and VIPER), it also allows to achieve these results with a single model and a single parameter setting. Our trained model and code are available at https://github.com/cv-stuttgart/MS_RAFT_plus .

While recent methods for motion and stereo estimation recover an unprecedented amount of details, such highly detailed structures are neither adequately reflected in the data of existing benchmarks nor their evaluation methodology. Hence, we introduce Spring - a large, high-resolution, high-detail, computer-generated benchmark for scene flow, optical flow, and stereo. Based on rendered scenes from the open-source Blender movie “Spring”, it provides photo-realistic HD datasets with state-of-the-art visual effects and ground truth training data. Furthermore, we provide a website to upload, analyze and compare results. Using a novel evaluation methodology based on a super-resolved UHD ground truth, our Spring benchmark can assess the quality of fine structures and provides further detailed performance statistics on different image regions. Regarding the number of ground truth frames, Spring is 60× larger than the only scene flow benchmark, KITTI 2015, and 15× larger than the well-established MPI Sintel optical flow benchmark. Initial results for recent methods on our benchmark show that estimating fine details is indeed challenging, as their accuracy leaves significant room for improvement. The Spring benchmark and the corresponding datasets are available at http://spring-benchmark.org.

Current adversarial attacks on motion estimation, or optical flow, optimize small per-pixel perturbations, which are unlikely to appear in the real world. In contrast, adverse weather conditions constitute a much more realistic threat scenario. Hence, in this work, we present a novel attack on motion estimation that exploits adversarially optimized particles to mimic weather effects like snowflakes, rain streaks or fog clouds. At the core of our attack framework is a differentiable particle rendering system that integrates particles (i) consistently over multiple time steps (ii) into the 3D space (iii) with a photo-realistic appearance. Through optimization, we obtain adversarial weather that significantly impacts the motion estimation. Surprisingly, methods that previously showed good robustness towards small per-pixel perturbations are particularly vulnerable to adversarial weather. At the same time, augmenting the training with non-optimized weather increases a method’s robustness towards weather effects and improves generalizability at almost no additional cost. Our code is available at https://github.com/cv-stuttgart/DistractingDownpour.

Recently, neural network for scene flow estimation show impressive results on automotive data such as the KITTI benchmark. However, despite of using sophisticated rigidity assumptions and parametrizations, such networks are typically limited to only two frame pairs which does not allow them to exploit temporal information. In our paper we address this shortcoming by proposing a novel multi-frame approach that considers an additional preceding stereo pair. To this end, we proceed in two steps: Firstly, building upon the recent RAFT-3D approach, we develop an improved two-frame baseline by incorporating an advanced stereo method. Secondly, and even more importantly, exploiting the specific modeling concepts of RAFT-3D, we propose a U-Net architecture that performs a fusion of forward and backward flow estimates and hence allows to integrate temporal information on demand. Experiments on the KITTI benchmark do not only show that the advantages of the improved baseline and the temporal fusion approach complement each other, they also demonstrate that the computed scene flow is highly accurate. More precisely, our approach ranks second overall and first for the even more challenging foreground objects, in total outperforming the original RAFT-3D method by more than 16%. Code is available at https://github.com/cv-stuttgart/M-FUSE.

Many classical and learning-based optical flow methods rely on hierarchical concepts to improve both accuracy and robustness. However, one of the currently most successful approaches – RAFT – hardly exploits such concepts. In this work, we show that multi-scale ideas are still valuable. More precisely, using RAFT as a baseline, we propose a novel multi-scale neural network that combines several hierarchical concepts within a single estimation framework. These concepts include (i) a partially shared coarse-to-fine architecture, (ii) multi-scale features, (iii) a hierarchical cost volume and (iv) a multi-scale multi-iteration loss. Experiments on MPI Sintel and KITTI clearly demonstrate the benefits of our approach. They show not only substantial improvements compared to RAFT, but also state-of-the-art results – in particular in non-occluded regions. Code will be available at https://github.com/cv-stuttgart/MS_RAFT.

We propose a new technique for visual analytics and annotation of long-term pervasive eye tracking data for which a combined analysis of gaze and egocentric video is necessary. Our approach enables two important tasks for such data for hour-long videos from individual participants: (1) efficient annotation and (2) direct interpretation of the results. Exemplary time spans can be selected by the user and are then used as a query that initiates a fuzzy search of similar time spans based on gaze and video features. In an iterative refinement loop, the query interface then provides suggestions for the importance of individual features to improve the search results. A multi-layered timeline visualization shows an overview of annotated time spans. We demonstrate the efficiency of our approach for analyzing activities in about seven hours of video in a case study and discuss feedback on our approach from novices and experts performing the annotation task.

Professional video editing tools can generate slow-motion video by interpolating frames from video recorded at a standard frame rate. Thereby the perceptual quality of such interpolated slow-motion videos strongly depends on the underlying interpolation techniques. We built a novel benchmark database that is specifically tailored for interpolated slow-motion videos (KoSMo-1k). It consists of 1,350 interpolated video sequences, from 30 different content sources, along with their subjective quality ratings from up to ten subjective comparisons per video pair. Moreover, we evaluated the performance of twelve existing full-reference (FR) image/video quality assessment (I/VQA) methods on the benchmark. In this way, we are able to show that specifically tailored quality assessment methods for interpolated slow-motion videos are needed, since the evaluated methods — despite their good performance on real-time video databases — do not give satisfying results when it comes to frame interpolation.

Current benchmarks for optical flow algorithms evaluate the estimation quality by comparing their predicted flow field with the ground truth, and additionally may compare interpolated frames, based on these predictions, with the correct frames from the actual image sequences. For the latter comparisons, objective measures such as mean square errors are applied. However, for applications like image interpolation, the expected user's quality of experience cannot be fully deduced from such simple quality measures. Therefore, we conducted a subjective quality assessment study by crowdsourcing for the interpolated images provided in one of the optical flow benchmarks, the Middlebury benchmark. We used paired comparisons with forced choice and reconstructed absolute quality scale values according to Thurstone's model using the classical least squares method. The results give rise to a re-ranking of 141 participating algorithms w.r.t. visual quality of interpolated frames mostly based on optical flow estimation. Our re-ranking result shows the necessity of visual quality assessment as another evaluation metric for optical flow and frame interpolation benchmarks.

In spite of significant advances in Shape from Shading (SfS) over the last years, it is still a challenging task to design SfS approaches that are flexible enough to handle a wide range of input scenes. In this paper, we address this lack of flexibility by proposing a novel model that extends the range of possible applications. To this end, we consider the class of modern perspective SfS models formulated via partial differential equations (PDEs). By combining a recent spherical surface parametrisation with the advanced non-Lambertian Oren-Nayar reflectance model, we obtain a robust approach that allows to deal with an arbitrary position of the light source while being able to handle rough surfaces and thus more realistic objects at the same time. To our knowledge, the resulting model is currently the most advanced and most flexible approach in the literature on PDE-based perspective SfS. Apart from deriving our model, we also show how the corresponding set of sophisticated Hamilton-Jacobi equations can be efficiently solved by a specifically tailored fast marching scheme. Experiments with medical real-world data demonstrate that our model works in practice and that is offers the desired flexibility.

Since two years there is a recent trend in optical flow estimation to improve the results of state-of-the-art variational methods by applying additional filtering steps such as median filters, bilateral filters, and non-local techniques. So far, however, the application of such filters has been restricted to two-frame optical flow methods. In this paper, we go beyond this two-frame case and investigate the usefulness of such filtering steps for multi-frame optical flow estimation. Thereby we consider both the application to single flow fields as well as the filtering of the entire spatio-temporal flow volume. In this context, we propose the use of a joint trilateral filter that processes all flow fields simultaneously while imposing consistency of joint flow structures at the same time. Evaluations on the Middlebury benchmark clearly demonstrate the success of our filtering strategy. Achieving rank 3, our method yields state-of-the art results and significantly outperforms the baseline method providing considerably sharper results.

Due to their improved capability to handle realistic illumination scenarios, nonLambertian reflectance models are becoming increasingly more popular in the Shape from Shading (SfS) community. One of these advanced models is the Oren-Nayar model which is particularly suited to handle rough surfaces. However, not only the proper selection of the model is important, also the validation of stable and efficient algorithms plays a fundamental role when it comes to the practical applicability. While there are many works dealing with such algorithms in the case of Lambertian SfS, no such analysis has been performed so far for the Oren-Nayar model. In our paper we address this problem and present an in-depth study for such an advanced SfS model. To this end, we investigate under which conditions, i.e. model parameters, the Fast Marching (FM) method can be applied – a method that is known to be one of the most efficient algorithms for solving the underlying partial differential equations of Hamilton-Jacobi type. In this context, we do not only perform a general investigation of the model using Osher’s criterion for verifying the suitability of the FM method. We also conduct a parameter dependent analysis that shows, that FM can safely be used for the model for a wide range of settings relevant for practical applications. Thus, for the first time, it becomes possible to theoretically justify the use of the FM method as solver for the Oren-Nayar model which has been applied so far on a purely empirical basis only. Numerical experiments demonstrate the validity of our theoretical analysis. They show a stable behaviour of the FM method for the predicted range of model parameters.

Recent progress in passive facial performance capture has shown impressively detailed results on highly articulated motion. However, most methods rely on complex multi-camera set-ups, controlled lighting or fiducial markers. This prevents them from being used in general environments, outdoor scenes, during live action on a film set, or by freelance animators and everyday users who want to capture their digital selves. In this paper, we therefore propose a lightweight passive facial performance capture approach that is able to reconstruct high-quality dynamic facial geometry from only a single pair of stereo cameras. Our method succeeds under uncontrolled and time-varying lighting, and also in outdoor scenes. Our approach builds upon and extends recent image-based scene flow computation, lighting estimation and shading-based refinement algorithms. It integrates them into a pipeline that is specifically tailored towards facial performance reconstruction from challenging binocular footage under uncontrolled lighting. In an experimental evaluation, the strong capabilities of our method become explicit: We achieve detailed and spatio-temporally coherent results for expressive facial motion in both indoor and outdoor scenes -- even from low quality input images recorded with a hand-held consumer stereo camera. We believe that our approach is the first to capture facial performances of such high quality from a single stereo rig and we demonstrate that it brings facial performance capture out of the studio, into the wild, and within the reach of everybody.

Despite the fact that temporal coherence is undeniably one of the key aspects when processing video data, this concept has hardly been exploited in recent optical flow methods. In this paper, we will present a novel parametrization for multi-frame optical flow computation that naturally enables us to embed the assumption of a temporally coherent spatial flow structure, as well as the assumption that the optical flow is smooth along motion trajectories. While the first assumption is realized by expanding spatial regularization over multiple frames, the second assumption is imposed by two novel first- and second-order trajectorial smoothness terms. With respect to the latter, we investigate an adaptive decision scheme that makes a local (per pixel) or global (per sequence) selection of the most appropriate model possible. Experiments show the clear superiority of our approach when compared to existing strategies for imposing temporal coherence. Moreover, we demonstrate the state-of-the-art performance of our method by achieving Top 3 results at the widely used Middlebury benchmark.

Abstract Despite their high popularity, common high dynamic range (HDR) methods are still limited in their practical applicability: They assume that the input images are perfectly aligned, which is often violated in practise. Our paper does not only free the user from this unrealistic limitation, but even turns the missing alignment into an advantage: By exploiting the multiple exposures, we can create a super‐resolution image. The alignment step is performed by a modern energy‐based optic flow approach that takes into account the varying exposure conditions. Moreover, it produces dense displacement fields with subpixel precision. As a consequence, our approach can handle arbitrary complex motion patterns, caused by severe camera shake and moving objects. Additionally, it benefits from several advantages over existing strategies: (i) It is robust under outliers (noise, occlusions, saturation problems) and allows for sharp discontinuities in the displacement field. (ii) The alignment step neither requires camera calibration nor knowledge of the exposure times. (iii) It can be efficiently implemented on CPU and GPU architectures. After the alignment is performed, we use the obtained subpixel accurate displacement fields as input for an energy‐based, joint super‐resolution and HDR (SR‐HDR) approach. It introduces robust data terms and anisotropic smoothness terms in the SR‐HDR literature. Our experiments with challenging real world data demonstrate that these novelties are pivotal for the favourable performance of our approach.

Abstract We introduce a new global approach for image dithering, stippling, screening and sampling. It is inspired by the physical principles of electrostatics. Repelling forces between equally charged particles create a homogeneous distribution in flat areas, while attracting forces from the image brightness values ensure a high approximation quality. Our model is transparent and uses only two intuitive parameters: One steers the granularity of our halftoning approach, and the other its regularity. We evaluate two versions of our algorithm: A discrete version for dithering that ties points to grid positions, as well as a continuous one which does not have this restriction, and can thus be used for stippling or sampling density functions. Our methods create very few visual artefacts, reveal favourable blue‐noise behaviour in the frequency domain, and have a lower approximation error under Gaussian convolution than state‐of‐the‐art methods.

Although variational methods are among the most accurate techniques for estimating the optical flow, they have not yet entered the field of real-time vision. Main reason is the great popularity of standard numerical schemes that are easy to implement, however, at the expense of being too slow for real-time performance. In our paper we address this problem in two ways: (i) we present an improved version of the highly accurate technique of Brox et al. (2004). Thereby we show that a separate robustification of the constancy assumptions is very useful, in particular if the I-norm is used as penalizer. As a result, a method is obtained that yields the lowest angular errors in the literature, (ii) We develop an efficient numerical scheme for the proposed approach that allows real-time performance for sequences of size 160 /spl times/ 720. To this end, we combine two hierarchical strategies: a coarse-to-fine warping strategy as implementation of a fixed point iteration for a non-convex optimisation problem and a nonlinear full multigrid method - a so called full approximation scheme (FAS) - for solving the highly nonlinear equation systems at each warping level. In the experimental section the advantage of the proposed approach becomes obvious: Outperforming standard numerical schemes by two orders of magnitude frame rates of six high quality flow fields per second are obtained on a 3.06 GHz Pentium4 PC.

This paper investigates the usefulness of bidirectional multigrid methods for variational optical flow computations. Although these numerical schemes are among the fastest methods for solving equation systems, they are rarely applied in the field of computer vision. We demonstrate how to employ those numerical methods for the treatment of variational optical flow formulations and show that the efficiency of this approach even allows for real-time performance on standard PCs. As a representative for variational optic flow methods, we consider the recently introduced combined local-global method. It can be considered as a noise-robust generalization of the Horn and Schunck technique. We present a decoupled, as well as a coupled, version of the classical Gauss-Seidel solver, and we develop several multgrid implementations based on a discretization coarse grid approximation. In contrast, with standard bidirectional multigrid algorithms, we take advantage of intergrid transfer operators that allow for nondyadic grid hierarchies. As a consequence, no restrictions concerning the image size or the number of traversed levels have to be imposed. In the experimental section, we juxtapose the developed multigrid schemes and demonstrate their superior performance when compared to unidirectional multgrid methods and nonhierachical solvers. For the well-known 316 x 252 Yosemite sequence, we succeeded in computing the complete set of dense flow fields in three quarters of a second on a 3.06-GHz Pentium4 PC. This corresponds to a frame rate of 18 flow fields per second which outperforms the widely-used Gauss-Seidel method by almost three orders of magnitude.

We present an approach to parallel variational optical-flow computation by using an arbitrary partition of the image plane and iteratively solving related local variational problems associated with each subdomain. The approach is particularly suited for implementations on PC clusters because interprocess communication is minimized by restricting the exchange of data to a lower dimensional interface. Our mathematical formulation supports various generalizations to linear/nonlinear convex variational approaches, three-dimensional image sequences, spatiotemporal regularization, and unstructured geometries and triangulations. Results concerning the effects of interface preconditioning, as well as runtime and communication volume measurements on a PC cluster, are presented. Our approach provides a major step toward real-time two-dimensional image processing using off-the-shelf PC hardware and facilitates the efficient application of variational approaches to large-scale image processing problems.