Recent advances in multimodal large language models (MLLMs) have demonstrated remarkable capabilities in vision-language tasks, yet they often struggle with vision-centric scenarios where precise visual focus is needed for accurate reasoning. In this paper, we introduce Argus to address these limitations with a new visual attention grounding mechanism. Our approach employs object-centric grounding as visual chain-of-thought signals, enabling more effective goal-conditioned visual attention during multimodal reasoning tasks. Evaluations on diverse benchmarks demonstrate that Argus excels in both multi-modal reasoning tasks and referring object grounding tasks. Extensive analysis further validates various design choices of Argus, and reveals the effectiveness of explicit language-guided visual region-of-interest engagement in MLLMs, highlighting the importance of advancing multimodal intelligence from a visual-centric perspective.

The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks.

We introduce Eagle2.5, a frontier vision-language model (VLM) for long-context multimodal learning. Our work addresses the challenges in long video comprehension and high-resolution image understanding, introducing a generalist framework for both tasks. The proposed training framework incorporates Automatic Degrade Sampling and Image Area Preservation, two techniques that preserve contextual integrity and visual details. The framework also includes numerous efficiency optimizations in the pipeline for long-context data training. Finally, we propose Eagle-Video-110K, a novel dataset that integrates both story-level and clip-level annotations, facilitating long-video understanding. Eagle2.5 demonstrates substantial improvements on long-context multimodal benchmarks, providing a robust solution to the limitations of existing VLMs. Notably, our best model Eagle2.5-8B achieves 72.4\% on Video-MME with 512 input frames, matching the results of top-tier commercial model such as GPT-4o and large-scale open-source models like Qwen2.5-VL-72B and InternVL2.5-78B.

Despite tremendous progress in generating high-quality images using diffusion models, synthesizing a sequence of animated frames that are both photorealistic and temporally coherent is still in its infancy. While off-the-shelf billion-scale datasets for image generation are available, collecting similar video data of the same scale is still challenging. Also, training a video diffusion model is computationally much more expensive than its image counterpart. In this work, we explore finetuning a pretrained image diffusion model with video data as a practical solution for the video synthesis task. We find that naively extending the image noise prior to video noise prior in video diffusion leads to sub-optimal performance. Our carefully designed video noise prior leads to substantially better performance. Extensive experimental validation shows that our model, Preserve Your Own COrrelation (PYoCo), attains SOTA zero-shot text-to-video results on the UCF-101 and MSR-VTT benchmarks. It also achieves SOTA video generation quality on the small-scale UCF-101 benchmark with a 10× smaller model using significantly less computation than the prior art. The project page is available at https://research.nvidia.com/labs/dir/pyoco/.

Partial convolution weights convolutions with binary masks and renormalizes on valid pixels. It was originally proposed for image inpainting task because a corrupted image processed by a standard convolutional often leads to artifacts. Therefore, binary masks are constructed that define the valid and corrupted pixels, so that partial convolution results are only calculated based on valid pixels. It has been also used for conditional image synthesis task, so that when a scene is generated, convolution results of an instance depend only on the feature values that belong to the same instance. One of the unexplored applications for partial convolution is padding which is a critical component of modern convolutional networks. Common padding schemes make strong assumptions about how the padded data should be extrapolated. We show that these padding schemes impair model accuracy, whereas partial convolution based padding provides consistent improvements across a range of tasks. In this article, we review partial convolution applications under one framework. We conduct a comprehensive study of the partial convolution based padding on a variety of computer vision tasks, including image classification, 3D-convolution-based action recognition, and semantic segmentation. Our results suggest that partial convolution-based padding shows promising improvements over strong baselines.

Humans can easily infer the underlying 3D geometry and texture of an object only from a single 2D image. Current computer vision methods can do this, too, but suffer from view generalization problems – the models inferred tend to make poor predictions of appearance in novel views. As for generalization problems in machine learning, the difficulty is balancing single-view accuracy (cf. training error; bias) with novel view accuracy (cf. test error; variance). We describe a class of models whose geometric rigidity is easily controlled to manage this tradeoff. We describe a cycle consistency loss that improves view generalization (roughly, a model from a generated view should predict the original view well). View generalization of textures requires that models share texture information, so a car seen from the back still has headlights because other cars have headlights. We describe a cycle consistency loss that encourages model textures to be aligned, so as to encourage sharing. We compare our method against the state-of-the-art method and show both qualitative and quantitative improvements.

We introduce DiscoBox, a novel framework that jointly learns instance segmentation and semantic correspondence using bounding box supervision. Specifically, we propose a self-ensembling framework where instance segmentation and semantic correspondence are jointly guided by a structured teacher in addition to the bounding box supervision. The teacher is a structured energy model incorporating a pairwise potential and a cross-image potential to model the pairwise pixel relationships both within and across the boxes. Minimizing the teacher energy simultaneously yields refined object masks and dense correspondences between intra-class objects, which are taken as pseudo-labels to supervise the task network and provide positive/negative correspondence pairs for dense contrastive learning. We show a symbiotic relationship where the two tasks mutually benefit from each other. Our best model achieves 37.9% AP on COCO instance segmentation, surpassing prior weakly supervised methods and is competitive to supervised methods. We also obtain state of the art weakly supervised results on PASCAL VOC12 and PF-PASCAL with real-time inference.

Generative Adversarial Networks (GANs) produce impressive results on unconditional image generation when powered with large-scale image datasets. Yet generated images are still easy to spot especially on datasets with high variance (e.g. bedroom, church). In this paper, we propose various improvements to further push the boundaries in image generation. Specifically, we propose a novel dual contrastive loss and show that, with this loss, discriminator learns more generalized and distinguishable representations to incentivize generation. In addition, we revisit attention and extensively experiment with different attention blocks in the generator. We find attention to be still an important module for successful image generation even though it was not used in the recent state-of-the-art models. Lastly, we study different attention architectures in the discriminator, and propose a reference attention mechanism. By combining the strengths of these remedies, we improve the compelling state-of-the-art Fréchet Inception Distance (FID) by at least 17.5% on several benchmark datasets. We obtain even more significant improvements on compositional synthetic scenes (up to 47.5% in FID).

Image segmentation and edge detection are both central problems in perceptual grouping. It is therefore interesting to study how these two tasks can be coupled to benefit each other. Indeed, segmentation can be easily transformed into contour edges to guide edge learning. However, the converse is nontrivial since general edges may not always form closed contours. In this paper, we propose a principled end-to-end framework for coupled edge and segmentation learning, where edges are leveraged as pairwise similarity cues to guide segmentation. At the core of our framework is a recurrent module termed as dynamic graph propagation (DGP) layer that performs message passing on dynamically constructed graphs. The layer uses learned gating to dynamically select neighbors for message passing using max-pooling. The output from message passing is further gated with an edge signal to refine segmentation. Experiments demonstrate that the proposed framework is able to let both tasks mutually improve each other. On Cityscapes validation, our best model achieves 83.7% mIoU in semantic segmentation and 78.7% maximum F-score in semantic edge detection. Our method also leads to improved zero-shot robustness on Cityscapes with natural corruptions (Cityscapes-C).

Conditional image synthesis for generating photorealistic images serves various applications for content editing to content generation. Previous conditional image synthesis algorithms mostly rely on semantic maps, and often fail in complex environments where multiple instances occlude each other. We propose a panoptic aware image synthesis network to generate high fidelity and photorealistic images conditioned on panoptic maps which unify semantic and instance information. To achieve this, we efficiently use panoptic maps in convolution and upsampling layers. We show that with the proposed changes to the generator, we can improve on the previous state-of-the-art methods by generating images in complex instance interaction environments in higher fidelity and tiny objects in more details. Furthermore, our proposed method also outperforms the previous state-of-the-art methods in metrics of mean IoU (Intersection over Union), and detAP (Detection Average Precision).

Synthesizing larger texture images from a smaller exemplar is an important task in graphics and vision. The conventional CNNs, recently adopted for synthesis, require to train and test on the same set of images and fail to generalize to unseen images. This is mainly because those CNNs fully rely on convolutional and upsampling layers that operate locally and not suitable for a task as global as texture synthesis. In this work, inspired by the repetitive nature of texture patterns, we find that texture synthesis can be viewed as (local) \textit{upsampling} in the Fast Fourier Transform (FFT) domain. However, FFT of natural images exhibits high dynamic range and lacks local correlations. Therefore, to train CNNs we design a framework to perform FFT upsampling in feature space using deformable convolutions. Such design allows our framework to generalize to unseen images, and synthesize textures in a single pass. Extensive evaluations confirm that our method achieves state-of-the-art performance both quantitatively and qualitatively.

Learning to synthesize high frame rate videos via interpolation requires large quantities of high frame rate training videos, which, however, are scarce, especially at high resolutions. Here, we propose unsupervised techniques to synthesize high frame rate videos directly from low frame rate videos using cycle consistency. For a triplet of consecutive frames, we optimize models to minimize the discrepancy between the center frame and its cycle reconstruction, obtained by interpolating back from interpolated intermediate frames. This simple unsupervised constraint alone achieves results comparable with supervision using the ground truth intermediate frames. We further introduce a pseudo supervised loss term that enforces the interpolated frames to be consistent with predictions of a pre-trained interpolation model. The pseudo supervised loss term, used together with cycle consistency, can effectively adapt a pre-trained model to a new target domain. With no additional data and in a completely unsupervised fashion, our techniques significantly improve pre-trained models on new target domains, increasing PSNR values from 32.84dB to 33.05dB on the Slowflow and from 31.82dB to 32.53dB on the Sintel evaluation datasets.

Inverse rendering aims to estimate physical attributes of a scene, e.g., reflectance, geometry, and lighting, from image(s). Inverse rendering has been studied primarily for single objects or with methods that solve for only one of the scene attributes. We propose the first learning based approach that jointly estimates albedo, normals, and lighting of an indoor scene from a single image. Our key contribution is the Residual Appearance Renderer (RAR), which can be trained to synthesize complex appearance effects (e.g., inter-reflection, cast shadows, near-field illumination, and realistic shading), which would be neglected otherwise. This enables us to perform self-supervised learning on real data using a reconstruction loss, based on re-synthesizing the input image from the estimated components. We finetune with real data after pretraining with synthetic data. To this end, we use physically-based rendering to create a large-scale synthetic dataset, named SUNCG-PBR, which is a significant improvement over prior datasets. Experimental results show that our approach outperforms state-of-the-art methods that estimate one or more scene attributes.

Video-to-video synthesis (vid2vid) aims at converting an input semantic video, such as videos of human poses or segmentation masks, to an output photorealistic video. While the state-of-the-art of vid2vid has advanced significantly, existing approaches share two major limitations. First, they are data-hungry. Numerous images of a target human subject or a scene are required for training. Second, a learned model has limited generalization capability. A pose-to-human vid2vid model can only synthesize poses of the single person in the training set. It does not generalize to other humans that are not in the training set. To address the limitations, we propose a few-shot vid2vid framework, which learns to synthesize videos of previously unseen subjects or scenes by leveraging few example images of the target at test time. Our model achieves this few-shot generalization capability via a novel network weight generation module utilizing an attention mechanism. We conduct extensive experimental validations with comparisons to strong baselines using several large-scale video datasets including human-dancing videos, talking-head videos, and street-scene videos. The experimental results verify the effectiveness of the proposed framework in addressing the two limitations of existing vid2vid approaches.

The ability to edit materials of objects in images is desirable by many content creators. However, this is an extremely challenging task as it requires to disentangle intrinsic physical properties of an image. We propose an end-to-end network architecture that replicates the forward image formation process to accomplish this task. Specifically, given a single image, the network first predicts intrinsic properties, i.e. shape, illumination, and material, which are then provided to a rendering layer. This layer performs in-network image synthesis, thereby enabling the network to understand the physics behind the image formation process. The proposed rendering layer is fully differentiable, supports both diffuse and specular materials, and thus can be applicable in a variety of problem settings. We demonstrate a rich set of visually plausible material editing examples and provide an extensive comparative study.

In this work, we propose a new type of visibility measurement named Continuous Visibility Feature (CVF). We say that a point q on the mesh is continuously visible from another point p if there exists a geodesic path connecting p and q that is entirely visible by p. In order to efficiently estimate the continuous visibility for all the vertices in a model, we propose two approaches that use specific CVF properties to avoid exhaustive visibility tests. CVF is then measured as the area of the continuously visible region. With this stronger visibility measure, we show that CVF better encodes the surface and part information of mesh than the tradition line-of-sight based visibility. For example, we show that existing segmentation algorithms can generate better segmentation results using CVF and its variants than using other visibility-based shape descriptors, such as shape diameter function. Similar to visibility and other mesh surface features, continuous visibility would have many applications.

Maintaining clearance, or distance from obstacles and sampling efficient enough configurations on the medial axises are a vital component for successful motion planning. Maintaining high clearance often creates safer paths for robots. Having bias for sampling on medial axis also offers higher possibility to find a path in complex environment where the feasible configuration space only occupies a small proportion of the whole space. Inspired by the similarity between medial axis and max-margin scheme in optimization, especially in Support Vector Machine, we propose a new method to quickly construct the medial axis for the motion planning environment both in low and high dimensional space. However, directly applying the SVM classification on the large volume of uniformly sampled configurations suffers from huge computation and the medial axis is usually not the real medial axis due to SVM's optimization function's tolerance to the mis-classification. Instead, we show a method that can quickly push any configuration to the medial axis by using the characteristics of the Max-Margin's optimization function. Experiments in low and high dimensional space and comparisons with other medial-axis motion planning algorithm are shown.

While techniques that segment shapes into visually meaningful parts have generated impressive results, these techniques also have only focused on relatively simple shapes, such as those composed of a single object either without holes or with few simple holes. In many applications, shapes created from images can contain many overlapping objects and holes. These holes may come from sensor noise, may have important parts of the shape or may be arbitrarily complex. These complexities that appear in real-world 2D shapes can pose grand challenges to the existing part segmentation methods. In this paper, we propose a new decomposition method, called Dual-space Decomposition that handles complex 2D shapes by recognizing the importance of holes and classifying holes as either topological noise or structurally important features. Our method creates a nearly convex decomposition of a given shape by segmenting both the polygon itself and its complementary. We compare our results to segmentation produced by nonexpert human subjects. Based on two evaluation methods, we show that this new decomposition method creates statistically similar to those produced by human subjects.