Modeling dynamic scenes through 4D Gaussians offers high visual fidelity and fast rendering speeds, but comes with significant storage overhead. Recent approaches mitigate this cost by aggressively reducing the number of Gaussians. However, this inevitably removes Gaussians essential for high-quality rendering, leading to severe degradation in dynamic regions. In this paper, we introduce a novel 4D anchor-based framework that tackles the storage cost in different perspective. Rather than reducing the number of Gaussians, our method retains a sufficient quantity to accurately model dynamic contents, while compressing them into compact, grid-aligned 4D anchor features. Each anchor is processed by an MLP to spawn a set of neural 4D Gaussians, which represent a local spatiotemporal region. We design these neural 4D Gaussians to capture temporal changes with minimal parameters, making them well-suited for the MLP-based spawning. Moreover, we introduce a dynamic-aware anchor growing strategy to effectively assign additional anchors to under-reconstructed dynamic regions. Our method adjusts the accumulated gradients with Gaussians' temporal coverage, significantly improving reconstruction quality in dynamic regions. Experimental results highlight that our method achieves state-of-the-art visual quality in dynamic regions, outperforming all baselines by a large margin with practical storage costs.

Recent advancements in online Video Instance Segmentation (VIS) methods show notable performance improvements across benchmarks. However, the leading methods in the tracking-by-detection paradigm often result in temporally inconsistent predictions at both instance-level and pixel-level that lead to visually unsatisfactory outcomes. To address these challenges, we propose RoCoVIS, a simple yet effective approach that integrates segmentation and tracking to provide consistent online VIS. Our approach is an end-to-end sequential learning where object queries are propagated through mask predictions, improving the accuracy of temporal instance mapping at the pixel level. Additionally, we propose a new label assignment criterion in harmony with our approach. We also examine the limitations and challenges presented by the current standard evaluation protocol (AP) and suggest adopting additional metrics, Tube-Boundary AP and AP_Pool. RoCoVIS demonstrates superior performance on challenging VIS benchmarks with a Swin-L backbone and shows competitive results when employing a ResNet-50 backbone. By employing Tube-Boundary AP and AP_Pool as metrics to measure mask accuracy and consistency, RoCoVIS outperforms its counterpart, GenVIS, on the HQ-YTVIS and VIPSeg.

In this paper, we propose LSRNA, a novel framework for higher-resolution (exceeding 1K) image generation using diffusion models by leveraging super-resolution directly in the latent space. Existing diffusion models struggle with scaling beyond their training resolutions, often leading to structural distortions or content repetition. Reference-based methods address the issues by upsampling a low-resolution reference to guide higher-resolution generation. However, they face significant challenges: upsampling in latent space often causes manifold deviation, which degrades output quality. On the other hand, upsampling in RGB space tends to produce overly smoothed out puts. To overcome these limitations, LSRNA combines Latent space Super-Resolution (LSR) for manifold alignment and Region-wise Noise Addition (RNA) to enhance high-frequency details. Our extensive experiments demonstrate that integrating LSRNA outperforms state-of-the-art reference-based methods across various resolutions and metrics, while showing the critical role of latent space up-sampling in preserving detail and sharpness. The code will be available at https://github.com/3587jjh/LSRNA.

We present Omni-RGPT, a multimodal large language model designed to facilitate region-level comprehension for both images and videos. To achieve consistent region representation across spatio-temporal dimensions, we introduce Token Mark, a set of tokens highlighting the target regions within the visual feature space. These tokens are directly embedded into spatial regions using region prompts (e.g., boxes or masks) and simultaneously incorporated into the text prompt to specify the target, establishing a direct connection between visual and text tokens. To further support robust video understanding without requiring tracklets, we introduce an auxiliary task that guides Token Mark by leveraging the consistency of the tokens, enabling stable region interpretation across the video. Additionally, we introduce a large-scale region-level video instruction dataset (RegVID300k). Omni-RGPT achieves state-of-the-art results on image and video-based commonsense reasoning benchmarks while showing strong performance in captioning and referring expression comprehension tasks.

Object compositing, the task of placing and harmonizing objects in images of diverse visual scenes, has become an important task in computer vision with the rise of generative models. However, existing datasets lack the diversity and scale required to comprehensively explore real-world scenarios. We introduce ORIDa (Object-centric Real-world Image Composition Dataset), a large-scale, real-captured dataset containing over 30,000 images featuring 200 unique objects, each of which is presented across varied positions and scenes. ORIDa has two types of data: factual-counterfactual sets and factual-only scenes. The factual-counterfactual sets consist of four factual images showing an object in different positions within a scene and a single counterfactual (or background) image of the scene without the object, resulting in five images per scene. The factual-only scenes include a single image containing an object in a specific context, expanding the variety of environments. To our knowledge, ORIDa is the first publicly available dataset with its scale and complexity for real-world image composition. Extensive analysis and experiments highlight the value of ORIDa as a resource for advancing further research in object compositing.

Constructing a compressed latent space through a variational autoencoder (VAE) is the key for efficient 3D diffusion models. This paper introduces COD-VAE that encodes 3D shapes into a COmpact set of 1D latent vectors without sacrificing quality. COD-VAE introduces a two-stage autoencoder scheme to improve compression and decoding efficiency. First, our encoder block progressively compresses point clouds into compact latent vectors via intermediate point patches. Second, our triplane-based decoder reconstructs dense triplanes from latent vectors instead of directly decoding neural fields, significantly reducing computational overhead of neural fields decoding. Finally, we propose uncertainty-guided token pruning, which allocates resources adaptively by skipping computations in simpler regions and improves the decoder efficiency. Experimental results demonstrate that COD-VAE achieves 16x compression compared to the baseline while maintaining quality. This enables 20.8x speedup in generation, highlighting that a large number of latent vectors is not a prerequisite for high-quality reconstruction and generation. The code is available at https://github.com/join16/COD-VAE.

Video large language models (LLMs) achieve strong video understanding by leveraging a large number of spatio-temporal tokens, but suffer from quadratic computational scaling with token count. To address this, we propose a training-free spatio-temporal token merging method, named STTM. Our key insight is to exploit local spatial and temporal redundancy in video data which has been overlooked in prior work. STTM first transforms each frame into multi-granular spatial tokens using a coarse-to-fine search over a quadtree structure, then performs directed pairwise merging across the temporal dimension. This decomposed merging approach outperforms existing token reduction methods across six video QA benchmarks. Notably, STTM achieves a 2$\times$ speed-up with only a 0.5% accuracy drop under a 50% token budget, and a 3$\times$ speed-up with just a 2% drop under a 30% budget. Moreover, STTM is query-agnostic, allowing KV cache reuse across different questions for the same video. The project page is available at https://www.jshyun.me/projects/sttm.

We introduce Hierarchical Streaming Video Understanding, a task that combines online temporal action localization with free-form description generation. Given the scarcity of datasets with hierarchical and fine-grained temporal annotations, we demonstrate that LLMs can effectively group atomic actions into higher-level events, enriching existing datasets. We then propose OpenHOUSE (Open-ended Hierarchical Online Understanding System for Events), which extends streaming action perception beyond action classification. OpenHOUSE features a specialized streaming module that accurately detects boundaries between closely adjacent actions, nearly doubling the performance of direct extensions of existing methods. We envision the future of streaming action perception in the integration of powerful generative models, with OpenHOUSE representing a key step in that direction.

Computational color constancy, or white balancing, is a key module in a camera's image signal processor (ISP) that corrects color casts from scene lighting. Because this operation occurs in the camera-specific raw color space, white balance algorithms must adapt to different cameras. This paper introduces a learning-based method for cross-camera color constancy that generalizes to new cameras without retraining. Our method leverages pre-calibrated color correction matrices (CCMs) available on ISPs that map the camera's raw color space to a standard space (e.g., CIE XYZ). Our method uses these CCMs to transform predefined illumination colors (i.e., along the Planckian locus) into the test camera's raw space. The mapped illuminants are encoded into a compact camera fingerprint embedding (CFE) that enables the network to adapt to unseen cameras. To prevent overfitting due to limited cameras and CCMs during training, we introduce a data augmentation technique that interpolates between cameras and their CCMs. Experimental results across multiple datasets and backbones show that our method achieves state-of-the-art cross-camera color constancy while remaining lightweight and relying only on data readily available in camera ISPs.

Recent advances in novel view synthesis (NVS) have enabled real-time rendering with 3D Gaussian Splatting (3DGS). However, existing methods struggle with artifacts and missing regions when rendering from viewpoints that deviate from the training trajectory, limiting seamless scene exploration. To address this, we propose a 3DGS-based pipeline that generates additional training views to enhance reconstruction. We introduce an information-gain-driven virtual camera placement strategy to maximize scene coverage, followed by video diffusion priors to refine rendered results. Fine-tuning 3D Gaussians with these enhanced views significantly improves reconstruction quality. To evaluate our method, we present Wild-Explore, a benchmark designed for challenging scene exploration. Experiments demonstrate that our approach outperforms existing 3DGS-based methods, enabling high-quality, artifact-free rendering from arbitrary viewpoints. https://exploregs.github.io

Scaling has been a major driver of recent advancements in deep learning. Numerous empirical studies have found that scaling laws often follow the power-law and proposed several variants of power-law functions to predict the scaling behavior at larger scales. However, existing methods mostly rely on point estimation and do not quantify uncertainty, which is crucial for real-world applications involving decision-making problems such as determining the expected performance improvements achievable by investing additional computational resources. In this work, we explore a Bayesian framework based on Prior-data Fitted Networks (PFNs) for neural scaling law extrapolation. Specifically, we design a prior distribution that enables the sampling of infinitely many synthetic functions resembling real-world neural scaling laws, allowing our PFN to meta-learn the extrapolation. We validate the effectiveness of our approach on real-world neural scaling laws, comparing it against both the existing point estimation methods and Bayesian approaches. Our method demonstrates superior performance, particularly in data-limited scenarios such as Bayesian active learning, underscoring its potential for reliable, uncertainty-aware extrapolation in practical applications.

The vocabulary size in temporal action localization (TAL) is limited by the scarcity of large-scale annotated datasets. To overcome this, recent works integrate vision-language models (VLMs), such as CLIP, for openvocabulary TAL (OV-TAL). However, despite the success of VLMs trained on extensive datasets, existing OV-TAL methods still rely on human-labeled TAL datasets of limited size to train action localizers, limiting their generalizability. In this paper, we explore the scalability of self-training with unlabeled YouTube videos for OV-TAL. Our approach consists of two stages: (1) a class-agnostic action localizer is trained on a human-labeled TAL dataset to generate pseudo-labels for unlabeled videos, and (2) the large-scale pseudo-labeled dataset is then used to train the localizer. Extensive experiments demonstrate that leveraging web-scale videos in self-training significantly enhances the generalizability of an action localizer. Additionally, we identify limitations in existing OV-TAL evaluation schemes and propose a new benchmark for thorough assessment. Finally, we showcase the TAL performance of the large multimodal model Gemini-1.5 on our new benchmark. Code is released at https://github.com/HYUNJS/STOV-TAL.

White balance (WB) algorithms in many commercial cameras assume single and uniform illumination, leading to undesirable results when multiple lighting sources with different chromaticities exist in the scene. Prior research on multi-illuminant WB typically predicts illumination at the pixel level without fully grasping the scene's actual lighting conditions, including the number and color of light sources. This often results in unnatural outcomes lacking in overall consistency. To handle this problem, we present a deep white balancing model that leverages the slot attention, where each slot is in charge of representing individual illuminants. This design enables the model to generate chromaticities and weight maps for individual illuminants, which are then fused to compose the final illumination map. Furthermore, we propose the centroid-matching loss, which regulates the activation of each slot based on the color range, thereby enhancing the model to separate illumination more effectively. Our method achieves the state-of-the-art performance on both single- and multi-illuminant WB benchmarks, and also offers additional information such as the number of illuminants in the scene and their chromaticity. This capability allows for illumination editing, an application not feasible with prior methods.

The handling of long videos with complex and occluded sequences has recently emerged as a new challenge in the video instance segmentation (VIS) community. However, existing methods have limitations in addressing this challenge. We argue that the biggest bottleneck in current approaches is the discrepancy between training and inference. To effectively bridge this gap, we propose a Generalized framework for VIS, namely GenVIS, that achieves state-of-the-art performance on challenging benchmarks without designing complicated architectures or requiring extra post-processing. The key contribution of GenVIS is the learning strategy, which includes a query-based training pipeline for sequential learning with a novel target label assignment. Additionally, we introduce a memory that effectively acquires information from previous states. Thanks to the new perspective, which focuses on building relationships between separate frames or clips, GenVIS can be flexibly executed in both online and semi-online manner. We evaluate our approach on popular VIS benchmarks, achieving state-of-the-art results on YouTube-VIS 2019/2021/2022 and Occluded VIS (OVIS). Notably, we greatly outperform the state-of-the-art on the long VIS benchmark (OVIS), improving 5.6 AP with ResNet-50 backbone. Code is available at https://github.com/miranheo/GenVIS.

Temporal Action Localization (TAL) methods typically operate on top of feature sequences from a frozen snippet encoder that is pretrained with the Trimmed Action Classification (TAC) tasks, resulting in a task discrepancy problem. While existing TAL methods mitigate this issue either by re-training the encoder with a pretext task or by end-to-end fine-tuning, they commonly require an overload of high memory and computation. In this work, we introduce Soft-Landing (SoLa) strategy, an efficient yet effective framework to bridge the transferability gap between the pretrained encoder and the downstream tasks by incorporating a light-weight neural network, i.e., a SoLa module, on top of the frozen encoder. We also propose an unsupervised training scheme for the SoLa module; it learns with inter-frame Similarity Matching that uses the frame interval as its supervisory signal, eliminating the need for temporal annotations. Experimental evaluation on various benchmarks for downstream TAL tasks shows that our method effectively alleviates the task discrepancy problem with remarkable computational efficiency.

Object-centric learning (OCL) aspires general and compositional understanding of scenes by representing a scene as a collection of object-centric representations. OCL has also been extended to multi-view image and video datasets to apply various data-driven inductive biases by utilizing geometric or temporal information in the multi-image data. Single-view images carry less information about how to disentangle a given scene than videos or multi-view images do. Hence, owing to the difficulty of applying inductive biases, OCL for single-view images remains challenging, resulting in inconsistent learning of object-centric representation. To this end, we introduce a novel OCL framework for single-view images, SLot Attention via SHepherding (SLASH), which consists of two simple- yet-effective modules on top of Slot Attention. The new modules, Attention Refining Kernel (ARK) and Intermediate Point Predictor and Encoder (IPPE), respectively, prevent slots from being distracted by the background noise and indicate locations for slots to focus on to facilitate learning of object-centric representation. We also propose a weak semi-supervision approach for OCL, whilst our proposed framework can be used without any assistant annotation during the inference. Experiments show that our proposed method enables consistent learning of object-centric representation and achieves strong performance across four datasets. Code is available at https://github.com/object-understanding/SLASH.

Online Action Detection (OAD) is the task of identifying actions in streaming videos without access to future frames. Much effort has been devoted to effectively capturing long-range dependencies, with transformers receiving the spotlight for their ability to capture long-range temporal structures. In contrast, RNNs have received less attention lately, due to their lower performance compared to recent methods that utilize transformers. In this paper, we investigate the underlying reasons for the inferior performance of RNNs compared to transformer-based algorithms. Our findings indicate that the discrepancy between training and inference is the primary hindrance to the effective training of RNNs. To address this, we propose applying non-uniform weights to the loss computed at each time step, which allows the RNN model to learn from the predictions made in an environment that better resembles the inference stage. Extensive experiments on three benchmark datasets, THUMOS, TVSeries, and FineAction demonstrate that a minimal RNN-based model trained with the proposed methodology performs equally or better than the existing best methods with a significant increase in efficiency. The code is available at https://github.com/jbistanbul/MiniROAD.

For online video instance segmentation (VIS), fully utilizing the information from previous frames in an efficient manner is essential for real-time applications. Most previous methods follow a two-stage approach requiring additional computations such as RPN and RoIAlign, and do not fully exploit the available information in the video for all subtasks in VIS. In this paper, we propose a novel single-stage framework for online VIS built based on the grid structured feature representation. The grid-based features allow us to employ fully convolutional networks for real-time processing, and also to easily reuse and share features within different components. We also introduce cooperatively operating modules that aggregate information from available frames, in order to enrich the features for all subtasks in VIS. Our design fully takes advantage of previous information in a grid form for all tasks in VIS in an efficient way, and we achieved the new state-of-the-art accuracy (38.6 AP and 36.9 AP) and speed (40.0 FPS) on YouTube-VIS 2019 and 2021 datasets among online VIS methods. The code is available at https://github.com/SuHoHan95/VISOLQ

Recently, both long-tailed recognition and object tracking have made great advances individually. TAO benchmark presented a mixture of the two, long-tailed object tracking, in order to further reflect the aspect of the real-world. To date, existing solutions have adopted detectors showing robustness in long-tailed distributions, which derive per-frame results. Then, they used tracking algorithms that combine the temporally independent detections to finalize tracklets. However, as the approaches did not take temporal changes in scenes into account, inconsistent classification results in videos led to low overall performance. In this paper, we present a set classifier that improves accuracy of classifying tracklets by aggregating information from multiple viewpoints contained in a tracklet. To cope with sparse annotations in videos, we further propose augmentation of tracklets that can maximize data efficiency. The set classifier is plug-and-playable to existing object trackers, and highly improves the performance of long-tailed object tracking. By simply attaching our method to QDTrack on top of ResNet-101, we achieve the new state-of-the-art, 19.9% and 15.7%$TrackAP_{50}$on TAO validation and test sets, respectively. Our code is available at this link11https://github.com/sukjunhwang/setclassifier.

Generic Event Boundary Detection (GEBD) is a newly suggested video understanding task that aims to find one level deeper semantic boundaries of events. Bridging the gap between natural human perception and video understanding, it has various potential applications, including interpretable and semantically valid video parsing. Still at an early development stage, existing GEBD solvers are simple extensions of relevant video understanding tasks, disregarding GEBD's distinctive characteristics. In this paper, we propose a novel framework for unsupervised/supervised GEBD, by using the Temporal Self-similarity Matrix (TSM) as the video representation. The new Recursive TSM Parsing (RTP) algorithm exploits local diagonal patterns in TSM to detect boundaries, and it is combined with the Boundary Contrastive (BoCo) loss to train our encoder to generate more informative TSMs. Our framework can be applied to both unsupervised and supervised settings, with both achieving state-of-the-art performance by a huge margin in GEBD benchmark. Especially, our unsupervised method outperforms the previous state-of-the-art “supervised” model, implying its exceptional efficacy.

We introduce a novel paradigm for offline Video Instance Segmentation (VIS), based on the hypothesis that explicit object-oriented information can be a strong clue for understanding the context of the entire sequence. To this end, we propose VITA, a simple structure built on top of an off-the-shelf Transformer-based image instance segmentation model. Specifically, we use an image object detector as a means of distilling object-specific contexts into object tokens. VITA accomplishes video-level understanding by associating frame-level object tokens without using spatio-temporal backbone features. By effectively building relationships between objects using the condensed information, VITA achieves the state-of-the-art on VIS benchmarks with a ResNet-50 backbone: 49.8 AP, 45.7 AP on YouTube-VIS 2019 & 2021, and 19.6 AP on OVIS. Moreover, thanks to its object token-based structure that is disjoint from the backbone features, VITA shows several practical advantages that previous offline VIS methods have not explored - handling long and high-resolution videos with a common GPU, and freezing a frame-level detector trained on image domain. Code is available at the link.

Commercial adoption of automatic music composition requires the capability of generating diverse and high-quality music suitable for the desired context (e.g., music for romantic movies, action games, restaurants, etc.). In this paper, we introduce combinatorial music generation, a new task to create varying background music based on given conditions. Combinatorial music generation creates short samples of music with rich musical metadata, and combines them to produce a complete music. In addition, we introduce ComMU, the first symbolic music dataset consisting of short music samples and their corresponding 12 musical metadata for combinatorial music generation. Notable properties of ComMU are that (1) dataset is manually constructed by professional composers with an objective guideline that induces regularity, and (2) it has 12 musical metadata that embraces composers' intentions. Our results show that we can generate diverse high-quality music only with metadata, and that our unique metadata such as track-role and extended chord quality improves the capacity of the automatic composition. We highly recommend watching our video before reading the paper (https://pozalabs.github.io/ComMU/).

Dense Interspecies Face Embedding (DIFE) is a new direction for understanding faces of various animals by extracting common features among animal faces including human face. There are three main obstacles for interspecies face understanding: (1) lack of animal data compared to human, (2) ambiguous connection between faces of various animals, and (3) extreme shape and style variance. To cope with the lack of data, we utilize multi-teacher knowledge distillation of CSE and StyleGAN2 requiring no additional data or label. Then we synthesize pseudo pair images through the latent space exploration of StyleGAN2 to find implicit associations between different animal faces. Finally, we introduce the semantic matching loss to overcome the problem of extreme shape differences between species. To quantitatively evaluate our method over possible previous methodologies like unsupervised keypoint detection, we perform interspecies facial keypoint transfer on MAFL and AP-10K. Furthermore, the results of other applications like interspecies face image manipulation and dense keypoint transfer are provided. The code is available at https://github.com/kingsj0405/dife.

Panoptic segmentation, which is a novel task of unifying instance segmentation and semantic segmentation, has attracted a lot of attention lately. However, most of the previous methods are composed of multiple pathways with each pathway specialized to a designated segmentation task. In this paper, we propose to resolve panoptic segmentation in single-shot by integrating the execution flows. With the integrated pathway, a unified feature map called Panoptic-Feature is generated, which includes the information of both things and stuffs. Panoptic-Feature becomes more sophisticated by auxiliary problems that guide to cluster pixels that belong to the same instance and differentiate between objects of different classes. A collection of convolutional filters, where each filter represents either a thing or stuff, is applied to Panoptic-Feature at once, materializing the single-shot panoptic segmentation. Taking the advantages of both top-down and bottom-up approaches, our method, named SPINet, enjoys high efficiency and accuracy on major panoptic segmentation benchmarks: COCO and Cityscapes.

We propose a novel and unified solution for user-guided video object segmentation tasks. In this work, we consider two scenarios of user-guided segmentation: semi-supervised and interactive segmentation. Due to the nature of the problem, available cues - video frame(s) with object masks (or scribbles) - become richer with the intermediate predictions (or additional user inputs). However, the existing methods make it impossible to fully exploit this rich source of information. We resolve the issue by leveraging memory networks and learning to read relevant information from all available sources. In the semi-supervised scenario, the previous frames with object masks form an external memory, and the current frame as the query is segmented using the information in the memory. Similarly, to work with user interactions, the frames that are given user inputs form the memory that guides segmentation. Internally, the query and the memory are densely matched in the feature space, covering all the space-time pixel locations in a feed-forward fashion. The abundant use of the guidance information allows us to better handle challenges such as appearance changes and occlusions. We validate our method on the latest benchmark sets and achieve state-of-the-art performance along with a fast runtime.

A number of super-resolution (SR) algorithms from interpolation to deep neural networks (DNN) have emerged to restore or create missing details of the input low-resolution image. As mobile devices and display hardware develops, the demand for practical SR technology has increased. Current state-of-the-art SR methods are based on DNNs for better quality. However, they are feasible when executed by using a parallel computing module (e.g. GPUs), and have been difficult to apply to general uses such as end-user software, smartphones, and televisions. To this end, we propose an efficient and practical approach for the SR by adopting look-up table (LUT). We train a deep SR network with a small receptive field and transfer the output values of the learned deep model to the LUT. At test time, we retrieve the precomputed HR output values from the LUT for query LR input pixels. The proposed method can be performed very quickly because it does not require a large number of floating point operations. Experimental results show the efficiency and the effectiveness of our method. Especially, our method runs faster while showing better quality compared to bicubic interpolation.

By the one-to-many nature of the super-resolution (SR) problem, a single low-resolution (LR) image can be mapped to many high-resolution (HR) images. However, learning based SR algorithms are trained to map an LR image to the corresponding ground truth (GT) HR image in the training dataset. The training loss will increase and penalize the algorithm when the output does not exactly match the GT target, even when the outputs are mathematically valid candidates according to the SR framework. This becomes more problematic for the blind SR, as diverse unknown blur kernels exacerbate the ill-posedness of the problem. To this end, we propose a fundamentally different approach for the SR by introducing the concept of the adaptive target. The adaptive target is generated from the original GT target by a transformation to match the output of the SR network. The adaptive target provides an effective way for the SR algorithm to deal with the ill-posed nature of the SR, by providing the algorithm with the flexibility of accepting a variety of valid solutions. Experimental results show the effectiveness of our algorithm, especially for improving the perceptual quality of HR outputs.

In this paper, we propose a novel learning-based polygonal point set tracking method. Compared to existing video object segmentation (VOS) methods that propagate pixel-wise object mask information, we propagate a polygonal point set over frames. Specifically, the set is defined as a subset of points in the target contour, and our goal is to track corresponding points on the target contour. Those outputs enable us to apply various visual effects such as motion tracking, part deformation, and texture mapping. To this end, we propose a new method to track the corresponding points between frames by the global-local alignment with delicately designed losses and regularization terms. We also introduce a novel learning strategy using synthetic and VOS datasets that makes it possible to tackle the problem without developing the point correspondence dataset. Since the existing datasets are not suitable to validate our method, we build a new polygonal point set tracking dataset and demonstrate the superior performance of our method over the baselines and existing contour-based VOS methods. In addition, we present visual-effects applications of our method on part distortion and text mapping.

Temporal action localization has been one of the most popular tasks in video understanding, due to the importance of detecting action instances in videos. However, not much progress has been made on extending it to work in an on-line fashion, although many video related tasks can benefit by going online with the growing video streaming services. To this end, we introduce a new task called Online Temporal Action Localization (On-TAL), in which the goal is to immediately detect action instances from an untrimmed streaming video. The online setting makes the new task very challenging as the actionness decision for every frame has to be made without access to future frames and also because post-processing methods cannot be used to modify past action proposals. We propose a novel framework, Context-Aware Actionness Grouping (CAG) as a solution for On-TAL and train it with the imitation learning algorithm, which allows us to avoid sophisticated reward engineering. Evaluation of our work on THUMOS14 and Activitynet1.3 shows significant improvement over non-naive baselines, demonstrating the effectiveness of our approach. As a by-product, our method can also be used for the Online Detection of Action Start (ODAS), in which our method also outperforms previous state-of-the-art models.

We introduce a Large Scale Multi-Illuminant (LSMI) Dataset that contains 7,486 images, captured with three different cameras on more than 2,700 scenes with two or three illuminants. For each image in the dataset, the new dataset provides not only the pixel-wise ground truth illumination but also the chromaticity of each illuminant in the scene and the mixture ratio of illuminants per pixel. Images in our dataset are mostly captured with illuminants existing in the scene, and the ground truth illumination is computed by taking the difference between the images with different illumination combination. Therefore, our dataset captures natural composition in the real-world setting with wide field-of-view, providing more extensive dataset compared to existing datasets for multi-illumination white balance. As conventional single illuminant white balance algorithms cannot be directly applied, we also apply per-pixel DNN-based white balance algorithm and show its effectiveness against using patch-wise white balancing. We validate the benefits of our dataset through extensive analysis including a user-study, and expect the dataset to make meaningful contribution for future work in white balancing.

We propose a novel end-to-end solution for video instance segmentation (VIS) based on transformers. Recently, the per-clip pipeline shows superior performance over per-frame methods leveraging richer information from multiple frames. However, previous per-clip models require heavy computation and memory usage to achieve frame-to-frame communications, limiting practicality.In this work, we propose Inter-frame Communication Transformers (IFC), which significantly reduces the overhead for information-passing between frames by efficiently encoding the context within the input clip.Specifically, we propose to utilize concise memory tokens as a means of conveying information as well as summarizing each frame scene.The features of each frame are enriched and correlated with other frames through exchange of information between the precisely encoded memory tokens.We validate our method on the latest benchmark sets and achieved state-of-the-art performance (AP 42.6 on YouTube-VIS 2019 val set using the offline inference) while having a considerably fast runtime (89.4 FPS). Our method can also be applied to near-online inference for processing a video in real-time with only a small delay.The code is available at https://github.com/sukjunhwang/IFC

Time-lapse videos usually contain visually appealing content but are often difficult and costly to create. In this paper, we present an end-to-end solution to synthesize a time-lapse video from a single outdoor image using deep neural networks. Our key idea is to train a conditional generative adversarial network based on existing datasets of time-lapse videos and image sequences. We propose a multi-frame joint conditional generation framework to effectively learn the correlation between the illumination change of an outdoor scene and the time of the day. We further present a multi-domain training scheme for robust training of our generative models from two datasets with different distributions and missing timestamp labels. Compared to alternative time-lapse video synthesis algorithms, our method uses the timestamp as the control variable and does not require a reference video to guide the synthesis of the final output. We conduct ablation studies to validate our algorithm and compare with state-of-the-art techniques both qualitatively and quantitatively.

We present a deep learning method for the interactive video object segmentation. Our method is built upon two core operations, interaction and propagation, and each operation is conducted by Convolutional Neural Networks. The two networks are connected both internally and externally so that the networks are trained jointly and interact with each other to solve the complex video object segmentation problem. We propose a new multi-round training scheme for the interactive video object segmentation so that the networks can learn how to understand the user's intention and update incorrect estimations during the training. At the testing time, our method produces high-quality results and also runs fast enough to work with users interactively. We evaluated the proposed method quantitatively on the interactive track benchmark at the DAVIS Challenge 2018. We outperformed other competing methods by a significant margin in both the speed and the accuracy. We also demonstrated that our method works well with real user interactions.

We present a novel deep learning based algorithm for video inpainting. Video inpainting is a process of completing corrupted or missing regions in videos. Video inpainting has additional challenges compared to image inpainting due to the extra temporal information as well as the need for maintaining the temporal coherency. We propose a novel DNN-based framework called the Copy-and-Paste Networks for video inpainting that takes advantage of additional information in other frames of the video. The network is trained to copy corresponding contents in reference frames and paste them to fill the holes in the target frame. Our network also includes an alignment network that computes homographies between frames for the alignment, enabling the network to take information from more distant frames for robustness. Our method produces visually pleasing and temporally coherent results while running faster than the state-of-the-art optimization-based method. In addition, we extend our framework for enhancing over/under exposed frames in videos. Using this enhancement technique, we were able to significantly improve the lane detection accuracy on road videos.

We propose the onion-peel networks for video completion. Given a set of reference images and a target image with holes, our network fills the hole by referring the contents in the reference images. Our onion-peel network progressively fills the hole from the hole boundary enabling it to exploit richer contextual information for the missing regions every step. Given a sufficient number of recurrences, even a large hole can be inpainted successfully. To attend to the missing information visible in the reference images, we propose an asymmetric attention block that computes similarities between the hole boundary pixels in the target and the non-hole pixels in the references in a non-local manner. With our attention block, our network can have an unlimited spatial-temporal window size and fill the holes with globally coherent contents. In addition, our framework is applicable to the image completion guided by the reference images without any modification, which is difficult to do with the previous methods. We validate that our method produces visually pleasing image and video inpainting results in realistic test cases.

We propose a novel solution for semi-supervised video object segmentation. By the nature of the problem, available cues (e.g. video frame(s) with object masks) become richer with the intermediate predictions. However, the existing methods are unable to fully exploit this rich source of information. We resolve the issue by leveraging memory networks and learn to read relevant information from all available sources. In our framework, the past frames with object masks form an external memory, and the current frame as the query is segmented using the mask information in the memory. Specifically, the query and the memory are densely matched in the feature space, covering all the space-time pixel locations in a feed-forward fashion. Contrast to the previous approaches, the abundant use of the guidance information allows us to better handle the challenges such as appearance changes and occlussions. We validate our method on the latest benchmark sets and achieved the state-of-the-art performance (overall score of 79.4 on Youtube-VOS val set, J of 88.7 and 79.2 on DAVIS 2016/2017 val set respectively) while having a fast runtime (0.16 second/frame on DAVIS 2016 val set).

We propose a deep video prediction model conditioned on a single image and an action class. To generate future frames, we first detect keypoints of a moving object and predict future motion as a sequence of keypoints. The input image is then translated following the predicted keypoints sequence to compose future frames. Detecting the keypoints is central to our algorithm, and our method is trained to detect the keypoints of arbitrary objects in an unsupervised manner. Moreover, the detected keypoints of the original videos are used as pseudo-labels to learn the motion of objects. Experimental results show that our method is successfully applied to various datasets without the cost of labeling keypoints in videos. The detected keypoints are similar to human-annotated labels, and prediction results are more realistic compared to the previous methods.

Video super-resolution (VSR) has become even more important recently to provide high resolution (HR) contents for ultra high definition displays. While many deep learning based VSR methods have been proposed, most of them rely heavily on the accuracy of motion estimation and compensation. We introduce a fundamentally different framework for VSR in this paper. We propose a novel end-to-end deep neural network that generates dynamic upsampling filters and a residual image, which are computed depending on the local spatio-temporal neighborhood of each pixel to avoid explicit motion compensation. With our approach, an HR image is reconstructed directly from the input image using the dynamic upsampling filters, and the fine details are added through the computed residual. Our network with the help of a new data augmentation technique can generate much sharper HR videos with temporal consistency, compared with the previous methods. We also provide analysis of our network through extensive experiments to show how the network deals with motions implicitly.

We present an efficient method for the semi-supervised video object segmentation. Our method achieves accuracy competitive with state-of-the-art methods while running in a fraction of time compared to others. To this end, we propose a deep Siamese encoder-decoder network that is designed to take advantage of mask propagation and object detection while avoiding the weaknesses of both approaches. Our network, learned through a two-stage training process that exploits both synthetic and real data, works robustly without any online learning or post-processing. We validate our method on four benchmark sets that cover single and multiple object segmentation. On all the benchmark sets, our method shows comparable accuracy while having the order of magnitude faster runtime. We also provide extensive ablation and add-on studies to analyze and evaluate our framework.

Light field cameras capture both the spatial and the angular properties of light rays in space. Due to its property, one can compute the depth from light fields in uncontrolled lighting environments, which is a big advantage over active sensing devices. Depth computed from light fields can be used for many applications including 3D modelling and refocusing. However, light field images from hand-held cameras have very narrow baselines with noise, making the depth estimation difficult. Many approaches have been proposed to overcome these limitations for the light field depth estimation, but there is a clear trade-off between the accuracy and the speed in these methods. In this paper, we introduce a fast and accurate light field depth estimation method based on a fully-convolutional neural network. Our network is designed by considering the light field geometry and we also overcome the lack of training data by proposing light field specific data augmentation methods. We achieved the top rank in the HCI 4D Light Field Benchmark on most metrics, and we also demonstrate the effectiveness of the proposed method on real-world light-field images.

This paper addresses the problem of manipulating images using natural language description. Our task aims to semantically modify visual attributes of an object in an image according to the text describing the new visual appearance. Although existing methods synthesize images having new attributes, they do not fully preserve text-irrelevant contents of the original image. In this paper, we propose the text-adaptive generative adversarial network (TAGAN) to generate semantically manipulated images while preserving text-irrelevant contents. The key to our method is the text-adaptive discriminator that creates word level local discriminators according to input text to classify fine-grained attributes independently. With this discriminator, the generator learns to generate images where only regions that correspond to the given text is modified. Experimental results show that our method outperforms existing methods on CUB and Oxford-102 datasets, and our results were mostly preferred on a user study. Extensive analysis shows that our method is able to effectively disentangle visual attributes and produce pleasing outputs.

An image is a very effective tool for conveying emotions. Many researchers have investigated emotions in images by using various features extracted from images. In this paper, we focus on two high-level features, the object and the background, and assume that the semantic information in images is a good cue for predicting emotions. An object is one of the most important elements that define an image, and we discover through experiments that there is a high correlation between the objects and emotions in images in most cases. Even with the same object, there may be slight differences in emotion due to different backgrounds, and we use the semantic information of the background to improve the prediction performance. By combining the different levels of features, we build an emotion-based feedforward deep neural network that produces the emotion values of a given image. The output emotion values in our framework are continuous values in two-dimensional space (valence and arousal), which are more effective than using a small number of emotion categories to describe emotions. Experiments confirm the effectiveness of our network in predicting the emotions of images.

We present a novel deep learning framework that models the scene dependent image processing inside cameras. Often called as the radiometric calibration, the process of recovering RAWimages from processed images (JPEG format in the sRGB color space) is essential for many computer vision tasks that rely on physically accurate radiance values. All previous works rely on the deterministic imaging model where the color transformation stays the same regardless of the scene and thus they can only be applied for images taken under the manual mode. In this paper, we propose a datadriven approach to learn the scene dependent and locally varying image processing inside cameras under the automode. Our method incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers. The results show that we can model the imaging pipeline of different cameras that operate under the automode accurately in both directions (from RAW to sRGB, from sRGB to RAW) and we show how we can apply our method to improve the performance of image deblurring.

We present a novel coded exposure video technique for multi-image motion deblurring. The key idea of this paper is to capture video frames with a set of complementary fluttering patterns, which enables us to preserve all spectrum bands of a latent image and recover a sharp latent image. To achieve this, we introduce an algorithm for generating a complementary set of binary sequences based on the modern communication theory and implement the coded exposure video system with an off-the-shelf machine vision camera. To demonstrate the effectiveness of our method, we provide in-depth analyses of the theoretical bounds and the spectral gains of our method and other state-of-the-art computational imaging approaches. We further show deblurring results on various challenging examples with quantitative and qualitative comparisons to other computational image capturing methods used for image deblurring, and show how our method can be applied for protecting privacy in videos.

Modelling and analyzing noise in images is a fundamental task in many computer vision systems. Traditionally, noise has been modelled per color channel assuming that the color channels are independent. Although the color channels can be considered as mutually independent in camera RAW images, signals from different color channels get mixed during the imaging process inside the camera due to gamut mapping, tone-mapping, and compression. We show the influence of the in-camera imaging pipeline on noise and propose a new noise model in the 3D RGB space to accounts for the color channel mix-ups. A data-driven approach for determining the parameters of the new noise model is introduced as well as its application to image denoising. The experiments show that our noise model represents the noise in regular JPEG images more accurately compared to the previous models and is advantageous in image denoising.

Capturing hyperspectral images requires expensive and specialized hardware that is not readily accessible to most users. Digital cameras, on the other hand, are significantly cheaper in comparison and can be easily purchased and used. In this paper, we present a framework for reconstructing hyperspectral images by using multiple consumer-level digital cameras. Our approach works by exploiting the different spectral sensitivities of different camera sensors. In particular, due to the differences in spectral sensitivities of the cameras, different cameras yield different RGB measurements for the same spectral signal. We introduce an algorithm that is able to combine and convert these different RGB measurements into a single hyperspectral image for both indoor and outdoor scenes. This camera-based approach allows hyperspectral imaging at a fraction of the cost of most existing hyperspectral hardware. We validate the accuracy of our reconstruction against ground truth hyperspectral images (using both synthetic and real cases) and show its usage on relighting applications.

In this paper, we present a novel multi-image motion deblurring method utilizing the coded exposure technique. The key idea of our work is to capture video frames with a set of complementary fluttering patterns to preserve spatial frequency details. We introduce an algorithm for generating a complementary set of binary sequences based on the modern communication theory and implement the coded exposure video system with an off-the-shelf machine vision camera. The effectiveness of our method is demonstrated on various challenging examples with quantitative and qualitative comparisons to other computational image capturing methods used for image deblurring.

This paper introduces a new color transfer method which is a process of transferring color of an image to match the color of another image of the same scene. The color of a scene may vary from image to image because the photographs are taken at different times, with different cameras, and under different camera settings. To solve for a full nonlinear and nonparametric color mapping in the 3D RGB color space, we propose a scattered point interpolation scheme using moving least squares and strengthen it with a probabilistic modeling of the color transfer in the 3D color space to deal with mis-alignments and noise. Experiments show the effectiveness of our method over previous color transfer methods both quantitatively and qualitatively. In addition, our framework can be applied for various instances of color transfer such as transferring color between different camera models, camera settings, and illumination conditions, as well as for video color transfers.

Abstract Typical high dynamic range (HDR) imaging approaches based on multiple images have difficulties in handling moving objects and camera shakes, suffering from the ghosting effect and the loss of sharpness in the output HDR image. While there exist a variety of solutions for resolving such limitations, most of the existing algorithms are susceptible to complex motions, saturation, and occlusions. In this paper, we propose an HDR imaging approach using the coded electronic shutter which can capture a scene with row‐wise varying exposures in a single image. Our approach enables a direct extension of the dynamic range of the captured image without using multiple images, by photometrically calibrating rows with different exposures. Due to the concurrent capture of multiple exposures, misalignments of moving objects are naturally avoided with significant reduction in the ghosting effect. To handle the issues with under‐/over‐exposure, noise, and blurs, we present a coherent HDR imaging process where the problems are resolved one by one at each step. Experimental results with real photographs, captured using a coded electronic shutter, demonstrate that our method produces a high quality HDR images without the ghosting and blur artifacts.

Finding a good binary sequence is critical in determining the performance of the coded exposure imaging, but previous methods mostly rely on a random search for finding the binary codes, which could easily fail to find good long sequences due to the exponentially growing search space. In this paper, we present a new computationally efficient algorithm for generating the binary sequence, which is especially well suited for longer sequences. We show that the concept of the low autocorrelation binary sequence that has been well exploited in the information theory community can be applied for generating the fluttering patterns of the shutter, propose a new measure of a good binary sequence, and present a new algorithm by modifying the Legendre sequence for the coded exposure imaging. Experiments using both synthetic and real data show that our new algorithm consistently generates better binary sequences for the coded exposure problem, yielding better deblurring and resolution enhancement results compared to the previous methods for generating the binary codes.

This paper investigates the role that nonlinear camera response functions (CRFs) have on image deblurring. We present a comprehensive study to analyze the effects of CRFs on motion deblurring. In particular, we show how nonlinear CRFs can cause a spatially invariant blur to behave as a spatially varying blur. We prove that such nonlinearity can cause large errors around edges when directly applying deconvolution to a motion blurred image without CRF correction. These errors are inevitable even with a known point spread function (PSF) and with state-of-the-art regularization-based deconvolution algorithms. In addition, we show how CRFs can adversely affect PSF estimation algorithms in the case of blind deconvolution. To help counter these effects, we introduce two methods to estimate the CRF directly from one or more blurred images when the PSF is known or unknown. Our experimental results on synthetic and real images validate our analysis and demonstrate the robustness and accuracy of our approaches.

This paper investigates the role that nonlinear camera response functions (CRFs) have on image deblurring. In particular, we show how nonlinear CRFs can cause a spatially invariant blur to behave as a spatially varying blur. This can result in noticeable ringing artifacts when deconvolution is applied even with a known point spread function (PSF). In addition, we show how CRFs can adversely affect PSF estimation algorithms in the case of blind deconvolution. To help counter these effects, we introduce two methods to estimate the CRF directly from one or more blurred images when the PSF is known or unknown. While not as accurate as conventional CRF estimation algorithms based on multiple exposures or calibration patterns, our approach is still quite effective in improving deblurring results in situations where the CRF is unknown.

Due to the convenience of taking pictures with various digital cameras and mobile devices, people often end up with multiple shots of the same scene with only slight variations. To enhance photo triaging, which is a very common photowork activity, we propose an effective and easy-to-use brush-and-drag interface that allows the user to interactively explore and compare photos within a broader scene context. First, we brush to mark an area of interest on a photo with our finger(s); our tailored segmentation engine automatically determines corresponding image elements among the photos. Then, we can drag the segmented elements from different photos across the screen to explore them simultaneously, and further perform simple finger gestures to interactively rank photos, select favorites for sharing, or to remove unwanted ones. This novel interaction method was implemented on a consumer-level tablet computer and demonstrated to offer effective interactions in a user study.

Abstract Group photos are one of the most common types of digital images found in personal image collections and on social networks. One typical post‐processing task for group photos is to produce a key or legend to identify the people in the photo. This is most often done using simple bounding boxes. A more professional approach is to create a picture legend that uses either a full or partial silhouette to identify the individuals. This paper introduces an efficient method for producing picture legends for group photos. Our approach combines face detection with human shape priors into an interactive selection framework to allow users to quickly segment the individuals in a group photo. Our results are better than those obtained by general selection tools and can be produced in a fraction of the time.

We present a study of in-camera image processing through an extensive analysis of more than 10,000 images from over 30 cameras. The goal of this work is to investigate if image values can be transformed to physically meaningful values, and if so, when and how this can be done. From our analysis, we found a major limitation of the imaging model employed in conventional radiometric calibration methods and propose a new in-camera imaging model that fits well with today's cameras. With the new model, we present associated calibration procedures that allow us to convert sRGB images back to their original CCD RAW responses in a manner that is significantly more accurate than any existing methods. Additionally, we show how this new imaging model can be used to build an image correction application that converts an sRGB input image captured with the wrong camera settings to an sRGB output image that would have been recorded under the correct settings of a specific camera.

This paper describes a method to construct seamless image mosaics of a panoramic scene containing two predominate planes: a distant back plane and a ground plane that sweeps out from the camera's location. While this type of panorama can be stitched when the camera is carefully rotated about its optical center, such ideal scene capture is hard to perform correctly. Existing techniques use a single homography per image to perform alignment followed by seam cutting or image blending to hide inevitable alignments artifacts. In this paper, we demonstrate how to use two homographies per image to produce a more seamless image. Specifically, our approach blends the homographies in the alignment procedure to perform a nonlinear warping. Once the images are geometrically stitched, they are further processed to blend seams and reduce curvilinear visual artifacts due to the nonlinear warping. As demonstrated in our paper, our procedure is able to produce results for this type of scene where current state-of-the-art techniques fail.

We present a study of radiometric calibration and the in-camera imaging process through an extensive analysis of more than 10,000 images from over 30 cameras. The goal is to investigate if image values can be transformed to physically meaningful values and if so, when and how this can be done. From our analysis, we show that the conventional radiometric model fits well for image pixels with low color saturation but begins to degrade as color saturation level increases. This is due to the color mapping process which includes gamut mapping in the in-camera processing that cannot be modeled with conventional methods. To this end, we introduce a new imaging model for radiometric calibration and present an effective calibration scheme that allows us to compensate for the nonlinear color correction to convert non-linear sRGB images to CCD RAW responses.

This paper presents an interactive visualization tool to study and analyze hyperspectral images (HSI) of historical documents. This work is part of a collaborative effort with the Nationaal Archief of the Netherlands (NAN) and Art Innovation, a manufacturer of hyperspectral imaging hardware designed for old and fragile documents. The NAN is actively capturing HSI of historical documents for use in a variety of tasks related to the analysis and management of archival collections, from ink and paper analysis to monitoring the effects of environmental aging. To assist their work, we have developed a comprehensive visualization tool that offers an assortment of visualization and analysis methods, including interactive spectral selection, spectral similarity analysis, time-varying data analysis and visualization, and selective spectral band fusion. This paper describes our visualization software and how it is used to facilitate the tasks needed by our collaborators. Evaluation feedback from our collaborators on how this tool benefits their work is included.

In this work we propose a dynamic scene model to provide information about the presence of salient motion in the scene, and that could be used for focusing the attention of a pan/tilt/zoom camera, or for background modeling purposes. Rather than proposing a set of saliency detectors, we define what we mean by salient motion, and propose a precise model for it. Detecting salient motion becomes equivalent to detecting a model change. We derive optimal online procedures to solve this problem, which enable a very fast implementation. Promising results show that our model can effectively detect salient motion even in severely cluttered scenes, and while a camera is panning and tilting.

The images of an outdoor scene collected over time are valuable in studying the scene appearance variation which can lead to novel applications and help enhance existing methods that were constrained to controlled environments. However, the images do not reflect the true appearance of the scene in many cases due to the radiometric properties of the camera : the radiometric response function and the changing exposure. We introduce a new algorithm to compute the radiometric response function and the exposure of images given a sequence of images of a static outdoor scene where the illumination is changing. We use groups of pixels with constant behaviors towards the illumination change for the response estimation and introduce a sinusoidal lighting variation model representing the daily motion of the sun to compute the exposures.

In many computer vision systems, it is assumed that the image brightness of a point directly reflects the scene radiance of the point. However, the assumption does not hold in most cases due to nonlinear camera response function, exposure changes, and vignetting. The effects of these factors are most visible in image mosaics and textures of 3D models where colors look inconsistent and notable boundaries exist. In this paper, we propose a full radiometric calibration algorithm that includes robust estimation of the radiometric response function, exposures, and vignetting. By decoupling the effect of vignetting from the response function estimation, we approach each process in a manner that is robust to noise and outliers. We verify our algorithm with both synthetic and real data which shows significant improvement compared to existing methods. We apply our estimation results to radiometrically align images for seamless mosaics and 3D model textures. We also use our method to create high dynamic range (HDR) mosaics which are more representative of the scene than normal mosaics.

To capture the full brightness range of natural scenes, cameras automatically adjust the exposure value which causes the brightness of scene points to change from frame to frame. Given such a video sequence, we introduce a new method for tracking features and estimating the radiometric response function of the camera and the exposure difference between frames simultaneously. We model the global and nonlinear process that is responsible for the changes in image brightness rather than adapting to the changes locally and linearly which makes our tracking more robust to the change in brightness. The radiometric response function and the exposure difference between frames are also estimated in the process which enables further video processing algorithms to deal with the varying brightness.