This paper presents a unified approach to understanding dynamic scenes from casual videos. Large pretrained vision foundation models, such as vision-language, video depth prediction, motion tracking, and segmentation models, offer promising capabilities. However, training a single model for comprehensive 4D understanding remains challenging. We introduce Uni4D, a multi-stage optimization framework that harnesses multiple pretrained models to advance dynamic 3D modeling, including static/dynamic reconstruction, camera pose estimation, and dense 3D motion tracking. Our results show state-of-the-art performance in dynamic 4D modeling with superior visual quality. Notably, Uni4D requires no retraining or fine-tuning, highlighting the effectiveness of repurposing visual foundation models for 4D understanding.
Temporal consistency is critical in video prediction to ensure that outputs are coherent and free of artifacts. Traditional methods, such as temporal attention and 3D convolution, may struggle with significant object motion and may not capture long-range temporal dependencies in dynamic scenes. To address this gap, we propose the Tracktention Layer, a novel architectural component that explicitly integrates motion information using point tracks, i.e., sequences of corresponding points across frames. By incorporating these motion cues, the Tracktention Layer enhances temporal alignment and effectively handles complex object motions, maintaining consistent feature representations over time. Our approach is computationally efficient and can be seamlessly integrated into existing models, such as Vision Transformers, with minimal modification. It can be used to upgrade image-only models to state-of-the-art video ones, sometimes outperforming models natively designed for video prediction. We demonstrate this on video depth prediction and video colorization, where models augmented with the Tracktention Layer exhibit significantly improved temporal consistency compared to baselines.
Image-event joint depth estimation methods leverage complementary modalities for robust perception, yet face challenges in generalizability stemming from two factors: 1) limited annotated image-event-depth datasets causing insufficient cross-modal supervision, and 2) inherent frequency mismatches between static images and dynamic event streams with distinct spatiotemporal patterns, leading to ineffective feature fusion. To address this dual challenge, we propose Frequency-decoupled Unified Self-supervised Encoder (FUSE) with two synergistic components: The Parameter-efficient Self-supervised Transfer (PST) establishes cross-modal knowledge transfer through latent space alignment with image foundation models, effectively mitigating data scarcity by enabling joint encoding without depth ground truth. Complementing this, we propose the Frequency-Decoupled Fusion module (FreDFuse) to explicitly decouple high-frequency edge features from low-frequency structural components, resolving modality-specific frequency mismatches through physics-aware fusion. This combined approach enables FUSE to construct a universal image-event encoder that only requires lightweight decoder adaptation for target datasets. Extensive experiments demonstrate state-of-the-art performance with 14% and 24.9% improvements in Abs.Rel on MVSEC and DENSE datasets. The framework exhibits remarkable zero-shot adaptability to challenging scenarios including extreme lighting and motion blur, significantly advancing real-world deployment capabilities. The source code for our method is publicly available at: https://github.com/sunpihai-up/FUSE
Self-supervised depth estimation from monocular cameras in diverse outdoor conditions, such as daytime, rain, and nighttime, is challenging due to the difficulty of learning universal representations and the severe lack of labeled real-world adverse data. Previous methods either rely on synthetic inputs and pseudo-depth labels or directly apply daytime strategies to adverse conditions, resulting in suboptimal results. In this paper, we present the first synthetic-to-real robust depth estimation framework, incorporating motion and structure priors to capture real-world knowledge effectively. In the synthetic adaptation, we transfer motion-structure knowledge inside cost volumes for better robust representation, using a frozen daytime model to train a depth estimator in synthetic adverse conditions. In the innovative real adaptation, which targets to fix synthetic-real gaps, models trained earlier identify the weather-insensitive regions with a designed consistency-reweighting strategy to emphasize valid pseudo-labels. We introduce a new regularization by gathering explicit depth distributions to constrain the model when facing real-world data. Experiments show that our method outperforms the state-of-the-art across diverse conditions in multi-frame and single-frame evaluations. We achieve improvements of 7.5% and 4.3% in AbsRel and RMSE on average for nuScenes and Robotcar datasets (daytime, nighttime, rain). In zero-shot evaluation of DrivingStereo (rain, fog), our method generalizes better than the previous ones.
In this paper, we introduce Semi-SD, a novel metric depth estimation framework tailored for surrounding cameras equipment in autonomous driving. In this work, the input data consists of adjacent surrounding frames and camera parameters. We propose a unified spatial-temporal-semantic fusion module to construct the visual fused features. Cross-attention components for surrounding cameras and adjacent frames are utilized to focus on metric scale information refinement and temporal feature matching. Building on this, we propose a pose estimation framework using surrounding cameras, their corresponding estimated depths, and extrinsic parameters, which effectively address the scale ambiguity in multi-camera setups. Moreover, semantic world model and monocular depth estimation world model are integrated to supervised the depth estimation, which improve the quality of depth estimation. We evaluate our algorithm on DDAD and nuScenes datasets, and the results demonstrate that our method achieves state-of-the-art performance in terms of surrounding camera based depth estimation quality. The source code will be available on https://github.com/xieyuser/Semi-SD.
We present Pow3r, a novel large 3D vision regression model that is highly versatile in the input modalities it accepts. Unlike previous feed-forward models that lack any mechanism to exploit known camera or scene priors at test time, Pow3r incorporates any combination of auxiliary information such as intrinsics, relative pose, dense or sparse depth, alongside input images, within a single network. Building upon the recent DUSt3R paradigm, a transformer-based architecture that leverages powerful pre-training, our lightweight and versatile conditioning acts as additional guidance for the network to predict more accurate estimates when auxiliary information is available. During training we feed the model with random subsets of modalities at each iteration, which enables the model to operate under different levels of known priors at test time. This in turn opens up new capabilities, such as performing inference in native image resolution, or point-cloud completion. Our experiments on 3D reconstruction, depth completion, multi-view depth prediction, multi-view stereo, and multi-view pose estimation tasks yield state-of-the-art results and confirm the effectiveness of Pow3r at exploiting all available information. The project webpage is https://europe.naverlabs.com/pow3r.
We propose PolyRad, a novel radar-guided depth estimation method that introduces polynomial fitting to transform scaleless depth predictions from pretrained monocular depth estimation (MDE) models into metric depth maps. Unlike existing approaches that rely on complex architectures or expensive sensors, our method is grounded in a simple yet fundamental insight: using polynomial coefficients predicted from cheap, ubiquitous radar data to adaptively adjust depth predictions non-uniformly across depth ranges. Although MDE models often infer reasonably accurate local depth structure within each object or local region, they may misalign these regions relative to one another, making a linear scale-and-shift transformation insufficient given three or more of these regions. In contrast, PolyRad generalizes beyond linear transformations and is able to correct such misalignments by introducing inflection points. Importantly, our polynomial fitting framework preserves structural consistency through a novel training objective that enforces monotonicity via first-derivative regularization. PolyRad achieves state-of-the-art performance on the nuScenes, ZJU-4DRadarCam, and View-of-Delft datasets, outperforming existing methods by 30.3% in MAE and 37.2% in RMSE.
Recent years have witnessed remarkable success of 3D Gaussian Splatting (3DGS) in novel view synthesis, surpassing prior differentiable rendering methods in both quality and efficiency. However, its training process suffers from coupled opacity-color optimization that frequently converges to local minima, producing floater artifacts that degrade visual fidelity. We present StableGS, a framework that eliminates floaters through cross-view depth consistency constraints while introducing a dual-opacity GS model to decouple geometry and material properties of translucent objects. To further enhance reconstruction quality in weakly-textured regions, we integrate DUSt3R depth estimation, significantly improving geometric stability. Our method fundamentally addresses 3DGS training instabilities, outperforming existing state-of-the-art methods across open-source datasets.
The integration of RGB and depth modalities significantly enhances the accuracy of segmenting complex indoor scenes, with depth data from RGB-D cameras playing a crucial role in this improvement. However, collecting an RGB-D dataset is more expensive than an RGB dataset due to the need for specialized depth sensors. Aligning depth and RGB images also poses challenges due to sensor positioning and issues like missing data and noise. In contrast, Pseudo Depth (PD) from high-precision depth estimation algorithms can eliminate the dependence on RGB-D sensors and alignment processes, as well as provide effective depth information and show significant potential in semantic segmentation. Therefore, to explore the practicality of utilizing pseudo depth instead of real depth for semantic segmentation, we design an RGB-PD segmentation pipeline to integrate RGB and pseudo depth and propose a Pseudo Depth Aggregation Module (PDAM) for fully exploiting the informative clues provided by the diverse pseudo depth maps. The PDAM aggregates multiple pseudo depth maps into a single modality, making it easily adaptable to other RGB-D segmentation methods. In addition, the pre-trained diffusion model serves as a strong feature extractor for RGB segmentation tasks, but multi-modal diffusion-based segmentation methods remain unexplored. Therefore, we present a Pseudo Depth Diffusion Model (PDDM) that adopts a large-scale text-image diffusion model as a feature extractor and a simple yet effective fusion strategy to integrate pseudo depth. To verify the applicability of pseudo depth and our PDDM, we perform extensive experiments on the NYUv2 and SUNRGB-D datasets. The experimental results demonstrate that pseudo depth can effectively enhance segmentation performance, and our PDDM achieves state-of-the-art performance, outperforming other methods by +6.98 mIoU on NYUv2 and +2.11 mIoU on SUNRGB-D.
Understanding the geometric and semantic properties of the scene is crucial in autonomous navigation and particularly challenging in the case of Unmanned Aerial Vehicle (UAV) navigation. Such information may be by obtained by estimating depth and semantic segmentation maps of the surrounding environment and for their practical use in autonomous navigation, the procedure must be performed as close to real-time as possible. In this paper, we leverage monocular cameras on aerial robots to predict depth and semantic maps in low-altitude unstructured environments. We propose a joint deep-learning architecture that can perform the two tasks accurately and rapidly, and validate its effectiveness on MidAir and Aeroscapes benchmark datasets. Our joint-architecture proves to be competitive or superior to the other single and joint architecture methods while performing its task fast predicting 20.2 FPS on a single NVIDIA quadro p5000 GPU and it has a low memory footprint. All codes for training and prediction can be found on this link: https://github.com/Malga-Vision/Co-SemDepth
Transparent and specular objects are frequently encountered in daily life, factories, and laboratories. However, due to the unique optical properties, the depth information on these objects is usually incomplete and inaccurate, which poses significant challenges for downstream robotics tasks. Therefore, it is crucial to accurately restore the depth information of transparent and specular objects. Previous depth completion methods for these objects usually use RGB information as an additional channel of the depth image to perform depth prediction. Due to the poor-texture characteristics of transparent and specular objects, these methods that rely heavily on color information tend to generate structure-less depth predictions. Moreover, these 2D methods cannot effectively explore the 3D structure hidden in the depth channel, resulting in depth ambiguity. To this end, we propose a geometry-aware assisted depth completion method for transparent and specular objects, which focuses on exploring the 3D structural cues of the scene. Specifically, besides extracting 2D features from RGB-D input, we back-project the input depth to a point cloud and build the 3D branch to extract hierarchical scene-level 3D structural features. To exploit 3D geometric information, we design several gated cross-modal fusion modules to effectively propagate multi-level 3D geometric features to the image branch. In addition, we propose an adaptive correlation aggregation strategy to appropriately assign 3D features to the corresponding 2D features. Extensive experiments on ClearGrasp, OOD, TransCG, and STD datasets show that our method outperforms other state-of-the-art methods. We further demonstrate that our method significantly enhances the performance of downstream robotic grasping tasks.
Despite the recent success of image-text contrastive models like CLIP and SigLIP, these models often struggle with vision-centric tasks that demand high-fidelity image understanding, such as counting, depth estimation, and fine-grained object recognition. These models, by performing language alignment, tend to prioritize high-level semantics over visual understanding, weakening their image understanding. On the other hand, vision-focused models are great at processing visual information but struggle to understand language, limiting their flexibility for language-driven tasks. In this work, we introduce TULIP, an open-source, drop-in replacement for existing CLIP-like models. Our method leverages generative data augmentation, enhanced image-image and text-text contrastive learning, and image/text reconstruction regularization to learn fine-grained visual features while preserving global semantic alignment. Our approach, scaling to over 1B parameters, outperforms existing state-of-the-art (SOTA) models across multiple benchmarks, establishing a new SOTA zero-shot performance on ImageNet-1K, delivering up to a $2\times$ enhancement over SigLIP on RxRx1 in linear probing for few-shot classification, and improving vision-language models, achieving over $3\times$ higher scores than SigLIP on MMVP. Our code/checkpoints are available at https://tulip-berkeley.github.io
Egocentric video-language pretraining has significantly advanced video representation learning. Humans perceive and interact with a fully 3D world, developing spatial awareness that extends beyond text-based understanding. However, most previous works learn from 1D text or 2D visual cues, such as bounding boxes, which inherently lack 3D understanding. To bridge this gap, we introduce EgoDTM, an Egocentric Depth- and Text-aware Model, jointly trained through large-scale 3D-aware video pretraining and video-text contrastive learning. EgoDTM incorporates a lightweight 3D-aware decoder to efficiently learn 3D-awareness from pseudo depth maps generated by depth estimation models. To further facilitate 3D-aware video pretraining, we enrich the original brief captions with hand-object visual cues by organically combining several foundation models. Extensive experiments demonstrate EgoDTM's superior performance across diverse downstream tasks, highlighting its superior 3D-aware visual understanding. Our code will be released at https://github.com/xuboshen/EgoDTM.
The increasing demand for high-accuracy depth estimation in autonomous driving and augmented reality applications necessitates advanced neural architectures capable of effectively leveraging multiple data modalities. In this context, we introduce the Unified Segmentation Attention Mechanism Network (USAM-Net), a novel convolutional neural network that integrates stereo image inputs with semantic segmentation maps and attention to enhance depth estimation performance. USAM-Net employs a dual-pathway architecture, which combines a pre-trained segmentation model (SAM) and a depth estimation model. The segmentation pathway preprocesses the stereo images to generate semantic masks, which are then concatenated with the stereo images as inputs to the depth estimation pathway. This integration allows the model to focus on important features such as object boundaries and surface textures which are crucial for accurate depth perception. Empirical evaluation on the DrivingStereo dataset demonstrates that USAM-Net achieves superior performance metrics, including a Global Difference (GD) of 3.61\% and an End-Point Error (EPE) of 0.88, outperforming traditional models such as CFNet, SegStereo, and iResNet. These results underscore the effectiveness of integrating segmentation information into stereo depth estimation tasks, highlighting the potential of USAM-Net in applications demanding high-precision depth data.
DUSt3R has recently shown that one can reduce many tasks in multi-view geometry, including estimating camera intrinsics and extrinsics, reconstructing the scene in 3D, and establishing image correspondences, to the prediction of a pair of viewpoint-invariant point maps, i.e., pixel-aligned point clouds defined in a common reference frame. This formulation is elegant and powerful, but unable to tackle dynamic scenes. To address this challenge, we introduce the concept of Dynamic Point Maps (DPM), extending standard point maps to support 4D tasks such as motion segmentation, scene flow estimation, 3D object tracking, and 2D correspondence. Our key intuition is that, when time is introduced, there are several possible spatial and time references that can be used to define the point maps. We identify a minimal subset of such combinations that can be regressed by a network to solve the sub tasks mentioned above. We train a DPM predictor on a mixture of synthetic and real data and evaluate it across diverse benchmarks for video depth prediction, dynamic point cloud reconstruction, 3D scene flow and object pose tracking, achieving state-of-the-art performance. Code, models and additional results are available at https://www.robots.ox.ac.uk/~vgg/research/dynamic-point-maps/.
Accurate 3D scene reconstruction is essential for numerous medical tasks. Given the challenges in obtaining ground truth data, there has been an increasing focus on self-supervised learning (SSL) for endoscopic depth estimation as a basis for scene reconstruction. While foundation models have shown remarkable progress in visual tasks, their direct application to the medical domain often leads to suboptimal results. However, the visual features from these models can still enhance endoscopic tasks, emphasizing the need for efficient adaptation strategies, which still lack exploration currently. In this paper, we introduce Endo3DAC, a unified framework for endoscopic scene reconstruction that efficiently adapts foundation models. We design an integrated network capable of simultaneously estimating depth maps, relative poses, and camera intrinsic parameters. By freezing the backbone foundation model and training only the specially designed Gated Dynamic Vector-Based Low-Rank Adaptation (GDV-LoRA) with separate decoder heads, Endo3DAC achieves superior depth and pose estimation while maintaining training efficiency. Additionally, we propose a 3D scene reconstruction pipeline that optimizes depth maps' scales, shifts, and a few parameters based on our integrated network. Extensive experiments across four endoscopic datasets demonstrate that Endo3DAC significantly outperforms other state-of-the-art methods while requiring fewer trainable parameters. To our knowledge, we are the first to utilize a single network that only requires surgical videos to perform both SSL depth estimation and scene reconstruction tasks. The code will be released upon acceptance.
In this paper, we propose Jasmine, the first Stable Diffusion (SD)-based self-supervised framework for monocular depth estimation, which effectively harnesses SD's visual priors to enhance the sharpness and generalization of unsupervised prediction. Previous SD-based methods are all supervised since adapting diffusion models for dense prediction requires high-precision supervision. In contrast, self-supervised reprojection suffers from inherent challenges (e.g., occlusions, texture-less regions, illumination variance), and the predictions exhibit blurs and artifacts that severely compromise SD's latent priors. To resolve this, we construct a novel surrogate task of hybrid image reconstruction. Without any additional supervision, it preserves the detail priors of SD models by reconstructing the images themselves while preventing depth estimation from degradation. Furthermore, to address the inherent misalignment between SD's scale and shift invariant estimation and self-supervised scale-invariant depth estimation, we build the Scale-Shift GRU. It not only bridges this distribution gap but also isolates the fine-grained texture of SD output against the interference of reprojection loss. Extensive experiments demonstrate that Jasmine achieves SoTA performance on the KITTI benchmark and exhibits superior zero-shot generalization across multiple datasets.
In this paper, we present a new method for multi-view geometric reconstruction. In recent years, large vision models have rapidly developed, performing excellently across various tasks and demonstrating remarkable generalization capabilities. Some works use large vision models for monocular depth estimation, which have been applied to facilitate multi-view reconstruction tasks in an indirect manner. Due to the ambiguity of the monocular depth estimation task, the estimated depth values are usually not accurate enough, limiting their utility in aiding multi-view reconstruction. We propose to incorporate SfM information, a strong multi-view prior, into the depth estimation process, thus enhancing the quality of depth prediction and enabling their direct application in multi-view geometric reconstruction. Experimental results on public real-world datasets show that our method significantly improves the quality of depth estimation compared to previous monocular depth estimation works. Additionally, we evaluate the reconstruction quality of our approach in various types of scenes including indoor, streetscape, and aerial views, surpassing state-of-the-art MVS methods. The code and supplementary materials are available at https://zju3dv.github.io/murre/ .
Recent multi-teacher distillation methods have unified the encoders of multiple foundation models into a single encoder, achieving competitive performance on core vision tasks like classification, segmentation, and depth estimation. This led us to ask: Could similar success be achieved when the pool of teachers also includes vision models specialized in diverse tasks across both 2D and 3D perception? In this paper, we define and investigate the problem of heterogeneous teacher distillation, or co-distillation, a challenging multi-teacher distillation scenario where teacher models vary significantly in both (a) their design objectives and (b) the data they were trained on. We explore data-sharing strategies and teacher-specific encoding, and introduce DUNE, a single encoder excelling in 2D vision, 3D understanding, and 3D human perception. Our model achieves performance comparable to that of its larger teachers, sometimes even outperforming them, on their respective tasks. Notably, DUNE surpasses MASt3R in Map-free Visual Relocalization with a much smaller encoder.
Multi-map Sparse Monocular visual Simultaneous Localization and Mapping applied to monocular endoscopic sequences has proven efficient to robustly recover tracking after the frequent losses in endoscopy due to motion blur, temporal occlusion, tools interaction or water jets. The sparse multi-maps are adequate for robust camera localization, however they are very poor for environment representation, they are noisy, with a high percentage of inaccurately reconstructed 3D points, including significant outliers, and more importantly with an unacceptable low density for clinical applications. We propose a method to remove outliers and densify the maps of the state of the art for sparse endoscopy multi-map CudaSIFT-SLAM. The NN LightDepth for up-to-scale depth dense predictions are aligned with the sparse CudaSIFT submaps by means of the robust to spurious LMedS. Our system mitigates the inherent scale ambiguity in monocular depth estimation while filtering outliers, leading to reliable densified 3D maps. We provide experimental evidence of accurate densified maps 4.15 mm RMS accuracy at affordable computing time in the C3VD phantom colon dataset. We report qualitative results on the real colonoscopy from the Endomapper dataset.