Video depth estimation lifts monocular video clips to 3D by inferring dense depth at every frame. Recent advances in single-image depth estimation, brought about by the rise of large foundation models and the use of synthetic training data, have fueled a renewed interest in video depth. However, naively applying a single-image depth estimator to every frame of a video disregards temporal continuity, which not only leads to flickering but may also break when camera motion causes sudden changes in depth range. An obvious and principled solution would be to build on top of video foundation models, but these come with their own limitations; including expensive training and inference, imperfect 3D consistency, and stitching routines for the fixed-length (short) outputs. We take a step back and demonstrate how to turn a single-image latent diffusion model (LDM) into a state-of-the-art video depth estimator. Our model, which we call RollingDepth, has two main ingredients: (i) a multi-frame depth estimator that is derived from a single-image LDM and maps very short video snippets (typically frame triplets) to depth snippets. (ii) a robust, optimization-based registration algorithm that optimally assembles depth snippets sampled at various different frame rates back into a consistent video. RollingDepth is able to efficiently handle long videos with hundreds of frames and delivers more accurate depth videos than both dedicated video depth estimators and high-performing single-frame models. Project page: rollingdepth.github.io.
360-degree images offer a significantly wider field of view compared to traditional pinhole cameras, enabling sparse sampling and dense 3D reconstruction in low-texture environments. This makes them crucial for applications in VR, AR, and related fields. However, the inherent distortion caused by the wide field of view affects feature extraction and matching, leading to geometric consistency issues in subsequent multi-view reconstruction. In this work, we propose 360Recon, an innovative MVS algorithm for ERP images. The proposed spherical feature extraction module effectively mitigates distortion effects, and by combining the constructed 3D cost volume with multi-scale enhanced features from ERP images, our approach achieves high-precision scene reconstruction while preserving local geometric consistency. Experimental results demonstrate that 360Recon achieves state-of-the-art performance and high efficiency in depth estimation and 3D reconstruction on existing public panoramic reconstruction datasets.
Self-supervised monocular depth estimation (MDE) has gained popularity for obtaining depth predictions directly from videos. However, these methods often produce scale invariant results, unless additional training signals are provided. Addressing this challenge, we introduce a novel self-supervised metric-scaled MDE model that requires only monocular video data and the camera's mounting position, both of which are readily available in modern vehicles. Our approach leverages planar-parallax geometry to reconstruct scene structure. The full pipeline consists of three main networks, a multi-frame network, a singleframe network, and a pose network. The multi-frame network processes sequential frames to estimate the structure of the static scene using planar-parallax geometry and the camera mounting position. Based on this reconstruction, it acts as a teacher, distilling knowledge such as scale information, masked drivable area, metric-scale depth for the static scene, and dynamic object mask to the singleframe network. It also aids the pose network in predicting a metric-scaled relative pose between two subsequent images. Our method achieved state-of-the-art results for the driving benchmark KITTI for metric-scaled depth prediction. Notably, it is one of the first methods to produce self-supervised metric-scaled depth prediction for the challenging Cityscapes dataset, demonstrating its effectiveness and versatility.
All-weather image restoration (AWIR) is crucial for reliable autonomous navigation under adverse weather conditions. AWIR models are trained to address a specific set of weather conditions such as fog, rain, and snow. But this causes them to often struggle with out-of-distribution (OoD) samples or unseen degradations which limits their effectiveness for real-world autonomous navigation. To overcome this issue, existing models must either be retrained or fine-tuned, both of which are inefficient and impractical, with retraining needing access to large datasets, and fine-tuning involving many parameters. In this paper, we propose using Low-Rank Adaptation (LoRA) to efficiently adapt a pre-trained all-weather model to novel weather restoration tasks. Furthermore, we observe that LoRA lowers the performance of the adapted model on the pre-trained restoration tasks. To address this issue, we introduce a LoRA-based fine-tuning method called LoRA-Align (LoRA-A) which seeks to align the singular vectors of the fine-tuned and pre-trained weight matrices using Singular Value Decomposition (SVD). This alignment helps preserve the model's knowledge of its original tasks while adapting it to unseen tasks. We show that images restored with LoRA and LoRA-A can be effectively used for computer vision tasks in autonomous navigation, such as semantic segmentation and depth estimation.
We propose SharpDepth, a novel approach to monocular metric depth estimation that combines the metric accuracy of discriminative depth estimation methods (e.g., Metric3D, UniDepth) with the fine-grained boundary sharpness typically achieved by generative methods (e.g., Marigold, Lotus). Traditional discriminative models trained on real-world data with sparse ground-truth depth can accurately predict metric depth but often produce over-smoothed or low-detail depth maps. Generative models, in contrast, are trained on synthetic data with dense ground truth, generating depth maps with sharp boundaries yet only providing relative depth with low accuracy. Our approach bridges these limitations by integrating metric accuracy with detailed boundary preservation, resulting in depth predictions that are both metrically precise and visually sharp. Our extensive zero-shot evaluations on standard depth estimation benchmarks confirm SharpDepth effectiveness, showing its ability to achieve both high depth accuracy and detailed representation, making it well-suited for applications requiring high-quality depth perception across diverse, real-world environments.
The generation of 3D scenes from user-specified conditions offers a promising avenue for alleviating the production burden in 3D applications. Previous studies required significant effort to realize the desired scene, owing to limited control conditions. We propose a method for controlling and generating 3D scenes under multimodal conditions using partial images, layout information represented in the top view, and text prompts. Combining these conditions to generate a 3D scene involves the following significant difficulties: (1) the creation of large datasets, (2) reflection on the interaction of multimodal conditions, and (3) domain dependence of the layout conditions. We decompose the process of 3D scene generation into 2D image generation from the given conditions and 3D scene generation from 2D images. 2D image generation is achieved by fine-tuning a pretrained text-to-image model with a small artificial dataset of partial images and layouts, and 3D scene generation is achieved by layout-conditioned depth estimation and neural radiance fields (NeRF), thereby avoiding the creation of large datasets. The use of a common representation of spatial information using 360-degree images allows for the consideration of multimodal condition interactions and reduces the domain dependence of the layout control. The experimental results qualitatively and quantitatively demonstrated that the proposed method can generate 3D scenes in diverse domains, from indoor to outdoor, according to multimodal conditions.
Despite progress in stereo depth estimation, omnidirectional imaging remains underexplored, mainly due to the lack of appropriate data. We introduce Helvipad, a real-world dataset for omnidirectional stereo depth estimation, featuring 40K video frames from video sequences across diverse environments, including crowded indoor and outdoor scenes with various lighting conditions. Collected using two 360{\deg} cameras in a top-bottom setup and a LiDAR sensor, the dataset includes accurate depth and disparity labels by projecting 3D point clouds onto equirectangular images. Additionally, we provide an augmented training set with an increased label density by using depth completion. We benchmark leading stereo depth estimation models for both standard and omnidirectional images. The results show that while recent stereo methods perform decently, a challenge persists in accurately estimating depth in omnidirectional imaging. To address this, we introduce necessary adaptations to stereo models, leading to improved performance.
Effectively manipulating articulated objects in household scenarios is a crucial step toward achieving general embodied artificial intelligence. Mainstream research in 3D vision has primarily focused on manipulation through depth perception and pose detection. However, in real-world environments, these methods often face challenges due to imperfect depth perception, such as with transparent lids and reflective handles. Moreover, they generally lack the diversity in part-based interactions required for flexible and adaptable manipulation. To address these challenges, we introduced a large-scale part-centric dataset for articulated object manipulation that features both photo-realistic material randomization and detailed annotations of part-oriented, scene-level actionable interaction poses. We evaluated the effectiveness of our dataset by integrating it with several state-of-the-art methods for depth estimation and interaction pose prediction. Additionally, we proposed a novel modular framework that delivers superior and robust performance for generalizable articulated object manipulation. Our extensive experiments demonstrate that our dataset significantly improves the performance of depth perception and actionable interaction pose prediction in both simulation and real-world scenarios. More information and demos can be found at: https://pku-epic.github.io/GAPartManip/.
Traditional monocular depth estimation suffers from inherent ambiguity and visual nuisance. We argue that language prior can enhance monocular depth estimation by leveraging the inductive bias learned during the text-to-image pre-training of diffusion models. The ability of these models to generate images that align with text indicates that they have learned the spatial relationships, size, and shape of specified objects, which can be applied to improve depth estimation. Thus, we propose PriorDiffusion, using a pre-trained text-to-image diffusion model that takes both images and corresponding text descriptions to infer affine-invariant depth through a denoising process. We also show that language prior enhances the model's perception of specific regions of images that users care about and describe. Simultaneously, language prior acts as a constraint to accelerate the convergence of both training and the inference diffusion trajectory. By training on HyperSim and Virtual KITTI, we achieve faster training convergence, fewer inference diffusion steps, and state-of-the-art zero-shot performance across NYUv2, KITTI, ETH3D, and ScanNet. Code will be released upon acceptance.
In this paper, we present DM-Calib, a diffusion-based approach for estimating pinhole camera intrinsic parameters from a single input image. Monocular camera calibration is essential for many 3D vision tasks. However, most existing methods depend on handcrafted assumptions or are constrained by limited training data, resulting in poor generalization across diverse real-world images. Recent advancements in stable diffusion models, trained on massive data, have shown the ability to generate high-quality images with varied characteristics. Emerging evidence indicates that these models implicitly capture the relationship between camera focal length and image content. Building on this insight, we explore how to leverage the powerful priors of diffusion models for monocular pinhole camera calibration. Specifically, we introduce a new image-based representation, termed Camera Image, which losslessly encodes the numerical camera intrinsics and integrates seamlessly with the diffusion framework. Using this representation, we reformulate the problem of estimating camera intrinsics as the generation of a dense Camera Image conditioned on an input image. By fine-tuning a stable diffusion model to generate a Camera Image from a single RGB input, we can extract camera intrinsics via a RANSAC operation. We further demonstrate that our monocular calibration method enhances performance across various 3D tasks, including zero-shot metric depth estimation, 3D metrology, pose estimation and sparse-view reconstruction. Extensive experiments on multiple public datasets show that our approach significantly outperforms baselines and provides broad benefits to 3D vision tasks. Code is available at https://github.com/JunyuanDeng/DM-Calib.
Recent progress in scene synthesis makes standalone SLAM systems purely based on optimizing hyperprimitives with a Rendering objective possible. However, the tracking performance still lacks behind traditional and end-to-end SLAM systems. An optimal trade-off between robustness, speed and accuracy has not yet been reached, especially for monocular video. In this paper, we introduce a SLAM system based on an end-to-end Tracker and extend it with a Renderer based on recent 3D Gaussian Splatting techniques. Our framework \textbf{DroidSplat} achieves both SotA tracking and rendering results on common SLAM benchmarks. We implemented multiple building blocks of modern SLAM systems to run in parallel, allowing for fast inference on common consumer GPU's. Recent progress in monocular depth prediction and camera calibration allows our system to achieve strong results even on in-the-wild data without known camera intrinsics. Code will be available at \url{https://github.com/ChenHoy/DROID-Splat}.
Large-scale pre-trained vision models are becoming increasingly prevalent, offering expressive and generalizable visual representations that benefit various downstream tasks. Recent studies on the emergent properties of these models have revealed their high-level geometric understanding, in particular in the context of depth perception. However, it remains unclear how depth perception arises in these models without explicit depth supervision provided during pre-training. To investigate this, we examine whether the monocular depth cues, similar to those used by the human visual system, emerge in these models. We introduce a new benchmark, DepthCues, designed to evaluate depth cue understanding, and present findings across 20 diverse and representative pre-trained vision models. Our analysis shows that human-like depth cues emerge in more recent larger models. We also explore enhancing depth perception in large vision models by fine-tuning on DepthCues, and find that even without dense depth supervision, this improves depth estimation. To support further research, our benchmark and evaluation code will be made publicly available for studying depth perception in vision models.
Recent advances in imitation learning have shown significant promise for robotic control and embodied intelligence. However, achieving robust generalization across diverse mounted camera observations remains a critical challenge. In this paper, we introduce a video-based spatial perception framework that leverages 3D spatial representations to address environmental variability, with a focus on handling lighting changes. Our approach integrates a novel image augmentation technique, AugBlender, with a state-of-the-art monocular depth estimation model trained on internet-scale data. Together, these components form a cohesive system designed to enhance robustness and adaptability in dynamic scenarios. Our results demonstrate that our approach significantly boosts the success rate across diverse camera exposures, where previous models experience performance collapse. Our findings highlight the potential of video-based spatial perception models in advancing robustness for end-to-end robotic learning, paving the way for scalable, low-cost solutions in embodied intelligence.
In this work, we introduce a generative approach for pose-free reconstruction of $360^{\circ}$ scenes from a limited number of uncalibrated 2D images. Pose-free scene reconstruction from incomplete, unposed observations is usually regularized with depth estimation or 3D foundational priors. While recent advances have enabled sparse-view reconstruction of unbounded scenes with known camera poses using diffusion priors, these methods rely on explicit camera embeddings for extrapolating unobserved regions. This reliance limits their application in pose-free settings, where view-specific data is only implicitly available. To address this, we propose an instruction-following RGBD diffusion model designed to inpaint missing details and remove artifacts in novel view renders and depth maps of a 3D scene. We also propose a novel confidence measure for Gaussian representations to allow for better detection of these artifacts. By progressively integrating these novel views in a Gaussian-SLAM-inspired process, we achieve a multi-view-consistent Gaussian representation. Evaluations on the MipNeRF360 dataset demonstrate that our method surpasses existing pose-free techniques and performs competitively with state-of-the-art posed reconstruction methods in complex $360^{\circ}$ scenes.
Given a video and a set of input object masks, an omnimatte method aims to decompose the video into semantically meaningful layers containing individual objects along with their associated effects, such as shadows and reflections. Existing omnimatte methods assume a static background or accurate pose and depth estimation and produce poor decompositions when these assumptions are violated. Furthermore, due to the lack of generative prior on natural videos, existing methods cannot complete dynamic occluded regions. We present a novel generative layered video decomposition framework to address the omnimatte problem. Our method does not assume a stationary scene or require camera pose or depth information and produces clean, complete layers, including convincing completions of occluded dynamic regions. Our core idea is to train a video diffusion model to identify and remove scene effects caused by a specific object. We show that this model can be finetuned from an existing video inpainting model with a small, carefully curated dataset, and demonstrate high-quality decompositions and editing results for a wide range of casually captured videos containing soft shadows, glossy reflections, splashing water, and more.
We introduce OneDiffusion, a versatile, large-scale diffusion model that seamlessly supports bidirectional image synthesis and understanding across diverse tasks. It enables conditional generation from inputs such as text, depth, pose, layout, and semantic maps, while also handling tasks like image deblurring, upscaling, and reverse processes such as depth estimation and segmentation. Additionally, OneDiffusion allows for multi-view generation, camera pose estimation, and instant personalization using sequential image inputs. Our model takes a straightforward yet effective approach by treating all tasks as frame sequences with varying noise scales during training, allowing any frame to act as a conditioning image at inference time. Our unified training framework removes the need for specialized architectures, supports scalable multi-task training, and adapts smoothly to any resolution, enhancing both generalization and scalability. Experimental results demonstrate competitive performance across tasks in both generation and prediction such as text-to-image, multiview generation, ID preservation, depth estimation and camera pose estimation despite relatively small training dataset. Our code and checkpoint are freely available at https://github.com/lehduong/OneDiffusion
Underwater environments pose significant challenges due to the selective absorption and scattering of light by water, which affects image clarity, contrast, and color fidelity. To overcome these, we introduce OceanLens, a method that models underwater image physics-encompassing both backscatter and attenuation-using neural networks. Our model incorporates adaptive backscatter and edge correction losses, specifically Sobel and LoG losses, to manage image variance and luminance, resulting in clearer and more accurate outputs. Additionally, we demonstrate the relevance of pre-trained monocular depth estimation models for generating underwater depth maps. Our evaluation compares the performance of various loss functions against state-of-the-art methods using the SeeThru dataset, revealing significant improvements. Specifically, we observe an average of 65% reduction in Grayscale Patch Mean Angular Error (GPMAE) and a 60% increase in the Underwater Image Quality Metric (UIQM) compared to the SeeThru and DeepSeeColor methods. Further, the results were improved with additional convolution layers that capture subtle image details more effectively with OceanLens. This architecture is validated on the UIEB dataset, with model performance assessed using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) metrics. OceanLens with multiple convolutional layers achieves up to 12-15% improvement in the SSIM.
This paper proposes a concise, elegant, and robust pipeline to estimate smooth camera trajectories and obtain dense point clouds for casual videos in the wild. Traditional frameworks, such as ParticleSfM~\cite{zhao2022particlesfm}, address this problem by sequentially computing the optical flow between adjacent frames to obtain point trajectories. They then remove dynamic trajectories through motion segmentation and perform global bundle adjustment. However, the process of estimating optical flow between two adjacent frames and chaining the matches can introduce cumulative errors. Additionally, motion segmentation combined with single-view depth estimation often faces challenges related to scale ambiguity. To tackle these challenges, we propose a dynamic-aware tracking any point (DATAP) method that leverages consistent video depth and point tracking. Specifically, our DATAP addresses these issues by estimating dense point tracking across the video sequence and predicting the visibility and dynamics of each point. By incorporating the consistent video depth prior, the performance of motion segmentation is enhanced. With the integration of DATAP, it becomes possible to estimate and optimize all camera poses simultaneously by performing global bundle adjustments for point tracking classified as static and visible, rather than relying on incremental camera registration. Extensive experiments on dynamic sequences, e.g., Sintel and TUM RGBD dynamic sequences, and on the wild video, e.g., DAVIS, demonstrate that the proposed method achieves state-of-the-art performance in terms of camera pose estimation even in complex dynamic challenge scenes.
Recent efforts in Gaussian-Splat-based Novel View Synthesis can achieve photorealistic rendering; however, such capability is limited in sparse-view scenarios due to sparse initialization and over-fitting floaters. Recent progress in depth estimation and alignment can provide dense point cloud with few views; however, the resulting pose accuracy is suboptimal. In this work, we present SPARS3R, which combines the advantages of accurate pose estimation from Structure-from-Motion and dense point cloud from depth estimation. To this end, SPARS3R first performs a Global Fusion Alignment process that maps a prior dense point cloud to a sparse point cloud from Structure-from-Motion based on triangulated correspondences. RANSAC is applied during this process to distinguish inliers and outliers. SPARS3R then performs a second, Semantic Outlier Alignment step, which extracts semantically coherent regions around the outliers and performs local alignment in these regions. Along with several improvements in the evaluation process, we demonstrate that SPARS3R can achieve photorealistic rendering with sparse images and significantly outperforms existing approaches.
Monocular geometric scene understanding combines panoptic segmentation and self-supervised depth estimation, focusing on real-time application in autonomous vehicles. We introduce MGNiceNet, a unified approach that uses a linked kernel formulation for panoptic segmentation and self-supervised depth estimation. MGNiceNet is based on the state-of-the-art real-time panoptic segmentation method RT-K-Net and extends the architecture to cover both panoptic segmentation and self-supervised monocular depth estimation. To this end, we introduce a tightly coupled self-supervised depth estimation predictor that explicitly uses information from the panoptic path for depth prediction. Furthermore, we introduce a panoptic-guided motion masking method to improve depth estimation without relying on video panoptic segmentation annotations. We evaluate our method on two popular autonomous driving datasets, Cityscapes and KITTI. Our model shows state-of-the-art results compared to other real-time methods and closes the gap to computationally more demanding methods. Source code and trained models are available at https://github.com/markusschoen/MGNiceNet.