Monocular depth estimation from a single image is an ill-posed problem for computer vision due to insufficient reliable cues as the prior knowledge. Besides the inter-frame supervision, namely stereo and adjacent frames, extensive prior information is available in the same frame. Reflections from specular surfaces, informative intra-frame priors, enable us to reformulate the ill-posed depth estimation task as a multi-view synthesis. This paper proposes the first self-supervision for deep-learning depth estimation on water scenes via intra-frame priors, known as reflection supervision and geometrical constraints. In the first stage, a water segmentation network is performed to separate the reflection components from the entire image. Next, we construct a self-supervised framework to predict the target appearance from reflections, perceived as other perspectives. The photometric re-projection error, incorporating SmoothL1 and a novel photometric adaptive SSIM, is formulated to optimize pose and depth estimation by aligning the transformed virtual depths and source ones. As a supplement, the water surface is determined from real and virtual camera positions, which complement the depth of the water area. Furthermore, to alleviate these laborious ground truth annotations, we introduce a large-scale water reflection scene (WRS) dataset rendered from Unreal Engine 4. Extensive experiments on the WRS dataset prove the feasibility of the proposed method compared to state-of-the-art depth estimation techniques.
Current monocular 3D scene reconstruction (3DR) works are either fully-supervised, or not generalizable, or implicit in 3D representation. We propose a novel framework - MonoSelfRecon that for the first time achieves explicit 3D mesh reconstruction for generalizable indoor scenes with monocular RGB views by purely self-supervision on voxel-SDF (signed distance function). MonoSelfRecon follows an Autoencoder-based architecture, decodes voxel-SDF and a generalizable Neural Radiance Field (NeRF), which is used to guide voxel-SDF in self-supervision. We propose novel self-supervised losses, which not only support pure self-supervision, but can be used together with supervised signals to further boost supervised training. Our experiments show that "MonoSelfRecon" trained in pure self-supervision outperforms current best self-supervised indoor depth estimation models and is comparable to 3DR models trained in fully supervision with depth annotations. MonoSelfRecon is not restricted by specific model design, which can be used to any models with voxel-SDF for purely self-supervised manner.
Road surface conditions, especially geometry profiles, enormously affect driving performance of autonomous vehicles. Vision-based online road reconstruction promisingly captures road information in advance. Existing solutions like monocular depth estimation and stereo matching suffer from modest performance. The recent technique of Bird's-Eye-View (BEV) perception provides immense potential to more reliable and accurate reconstruction. This paper uniformly proposes two simple yet effective models for road elevation reconstruction in BEV named RoadBEV-mono and RoadBEV-stereo, which estimate road elevation with monocular and stereo images, respectively. The former directly fits elevation values based on voxel features queried from image view, while the latter efficiently recognizes road elevation patterns based on BEV volume representing correlation between left and right voxel features. Insightful analyses reveal their consistence and difference with the perspective view. Experiments on real-world dataset verify the models' effectiveness and superiority. Elevation errors of RoadBEV-mono and RoadBEV-stereo achieve 1.83 cm and 0.50 cm, respectively. Our models are promising for practical road preview, providing essential information for promoting safety and comfort of autonomous vehicles. The code is released at https://github.com/ztsrxh/RoadBEV
Radar and camera fusion yields robustness in perception tasks by leveraging the strength of both sensors. The typical extracted radar point cloud is 2D without height information due to insufficient antennas along the elevation axis, which challenges the network performance. This work introduces a learning-based approach to infer the height of radar points associated with 3D objects. A novel robust regression loss is introduced to address the sparse target challenge. In addition, a multi-task training strategy is employed, emphasizing important features. The average radar absolute height error decreases from 1.69 to 0.25 meters compared to the state-of-the-art height extension method. The estimated target height values are used to preprocess and enrich radar data for downstream perception tasks. Integrating this refined radar information further enhances the performance of existing radar camera fusion models for object detection and depth estimation tasks.
Dense scene reconstruction for photo-realistic view synthesis has various applications, such as VR/AR, autonomous vehicles. However, most existing methods have difficulties in large-scale scenes due to three core challenges: \textit{(a) inaccurate depth input.} Accurate depth input is impossible to get in real-world large-scale scenes. \textit{(b) inaccurate pose estimation.} Most existing approaches rely on accurate pre-estimated camera poses. \textit{(c) insufficient scene representation capability.} A single global radiance field lacks the capacity to effectively scale to large-scale scenes. To this end, we propose an incremental joint learning framework, which can achieve accurate depth, pose estimation, and large-scale scene reconstruction. A vision transformer-based network is adopted as the backbone to enhance performance in scale information estimation. For pose estimation, a feature-metric bundle adjustment (FBA) method is designed for accurate and robust camera tracking in large-scale scenes. In terms of implicit scene representation, we propose an incremental scene representation method to construct the entire large-scale scene as multiple local radiance fields to enhance the scalability of 3D scene representation. Extended experiments have been conducted to demonstrate the effectiveness and accuracy of our method in depth estimation, pose estimation, and large-scale scene reconstruction.
Many established vision perception systems for autonomous driving scenarios ignore the influence of light conditions, one of the key elements for driving safety. To address this problem, we present HawkDrive, a novel perception system with hardware and software solutions. Hardware that utilizes stereo vision perception, which has been demonstrated to be a more reliable way of estimating depth information than monocular vision, is partnered with the edge computing device Nvidia Jetson Xavier AGX. Our software for low light enhancement, depth estimation, and semantic segmentation tasks, is a transformer-based neural network. Our software stack, which enables fast inference and noise reduction, is packaged into system modules in Robot Operating System 2 (ROS2). Our experimental results have shown that the proposed end-to-end system is effective in improving the depth estimation and semantic segmentation performance. Our dataset and codes will be released at https://github.com/ZionGo6/HawkDrive.
Phase-coded imaging is a computational imaging method designed to tackle tasks such as passive depth estimation and extended depth of field (EDOF) using depth cues inserted during image capture. Most of the current deep learning-based methods for depth estimation or all-in-focus imaging require a training dataset with high-quality depth maps and an optimal focus point at infinity for all-in-focus images. Such datasets are difficult to create, usually synthetic, and require external graphic programs. We propose a new method named "Deep Phase Coded Image Prior" (DPCIP) for jointly recovering the depth map and all-in-focus image from a coded-phase image using solely the captured image and the optical information of the imaging system. Our approach does not depend on any specific dataset and surpasses prior supervised techniques utilizing the same imaging system. This improvement is achieved through the utilization of a problem formulation based on implicit neural representation (INR) and deep image prior (DIP). Due to our zero-shot method, we overcome the barrier of acquiring accurate ground-truth data of depth maps and all-in-focus images for each new phase-coded system introduced. This allows focusing mainly on developing the imaging system, and not on ground-truth data collection.
Recovering the 3D scene geometry from a single view is a fundamental yet ill-posed problem in computer vision. While classical depth estimation methods infer only a 2.5D scene representation limited to the image plane, recent approaches based on radiance fields reconstruct a full 3D representation. However, these methods still struggle with occluded regions since inferring geometry without visual observation requires (i) semantic knowledge of the surroundings, and (ii) reasoning about spatial context. We propose KYN, a novel method for single-view scene reconstruction that reasons about semantic and spatial context to predict each point's density. We introduce a vision-language modulation module to enrich point features with fine-grained semantic information. We aggregate point representations across the scene through a language-guided spatial attention mechanism to yield per-point density predictions aware of the 3D semantic context. We show that KYN improves 3D shape recovery compared to predicting density for each 3D point in isolation. We achieve state-of-the-art results in scene and object reconstruction on KITTI-360, and show improved zero-shot generalization compared to prior work. Project page: https://ruili3.github.io/kyn.
We present MVD-Fusion: a method for single-view 3D inference via generative modeling of multi-view-consistent RGB-D images. While recent methods pursuing 3D inference advocate learning novel-view generative models, these generations are not 3D-consistent and require a distillation process to generate a 3D output. We instead cast the task of 3D inference as directly generating mutually-consistent multiple views and build on the insight that additionally inferring depth can provide a mechanism for enforcing this consistency. Specifically, we train a denoising diffusion model to generate multi-view RGB-D images given a single RGB input image and leverage the (intermediate noisy) depth estimates to obtain reprojection-based conditioning to maintain multi-view consistency. We train our model using large-scale synthetic dataset Obajverse as well as the real-world CO3D dataset comprising of generic camera viewpoints. We demonstrate that our approach can yield more accurate synthesis compared to recent state-of-the-art, including distillation-based 3D inference and prior multi-view generation methods. We also evaluate the geometry induced by our multi-view depth prediction and find that it yields a more accurate representation than other direct 3D inference approaches.
Monocular 3D object detection has attracted widespread attention due to its potential to accurately obtain object 3D localization from a single image at a low cost. Depth estimation is an essential but challenging subtask of monocular 3D object detection due to the ill-posedness of 2D to 3D mapping. Many methods explore multiple local depth clues such as object heights and keypoints and then formulate the object depth estimation as an ensemble of multiple depth predictions to mitigate the insufficiency of single-depth information. However, the errors of existing multiple depths tend to have the same sign, which hinders them from neutralizing each other and limits the overall accuracy of combined depth. To alleviate this problem, we propose to increase the complementarity of depths with two novel designs. First, we add a new depth prediction branch named complementary depth that utilizes global and efficient depth clues from the entire image rather than the local clues to reduce the correlation of depth predictions. Second, we propose to fully exploit the geometric relations between multiple depth clues to achieve complementarity in form. Benefiting from these designs, our method achieves higher complementarity. Experiments on the KITTI benchmark demonstrate that our method achieves state-of-the-art performance without introducing extra data. In addition, complementary depth can also be a lightweight and plug-and-play module to boost multiple existing monocular 3d object detectors. Code is available at https://github.com/elvintanhust/MonoCD.
Three-dimensional (3D) reconstruction from a single image is an ill-posed problem with inherent ambiguities, i.e. scale. Predicting a 3D scene from text description(s) is similarly ill-posed, i.e. spatial arrangements of objects described. We investigate the question of whether two inherently ambiguous modalities can be used in conjunction to produce metric-scaled reconstructions. To test this, we focus on monocular depth estimation, the problem of predicting a dense depth map from a single image, but with an additional text caption describing the scene. To this end, we begin by encoding the text caption as a mean and standard deviation; using a variational framework, we learn the distribution of the plausible metric reconstructions of 3D scenes corresponding to the text captions as a prior. To "select" a specific reconstruction or depth map, we encode the given image through a conditional sampler that samples from the latent space of the variational text encoder, which is then decoded to the output depth map. Our approach is trained alternatingly between the text and image branches: in one optimization step, we predict the mean and standard deviation from the text description and sample from a standard Gaussian, and in the other, we sample using a (image) conditional sampler. Once trained, we directly predict depth from the encoded text using the conditional sampler. We demonstrate our approach on indoor (NYUv2) and outdoor (KITTI) scenarios, where we show that language can consistently improve performance in both.
In self-supervised monocular depth estimation tasks, discrete disparity prediction has been proven to attain higher quality depth maps than common continuous methods. However, current discretization strategies often divide depth ranges of scenes into bins in a handcrafted and rigid manner, limiting model performance. In this paper, we propose a learnable module, Adaptive Discrete Disparity Volume (ADDV), which is capable of dynamically sensing depth distributions in different RGB images and generating adaptive bins for them. Without any extra supervision, this module can be integrated into existing CNN architectures, allowing networks to produce representative values for bins and a probability volume over them. Furthermore, we introduce novel training strategies - uniformizing and sharpening - through a loss term and temperature parameter, respectively, to provide regularizations under self-supervised conditions, preventing model degradation or collapse. Empirical results demonstrate that ADDV effectively processes global information, generating appropriate bins for various scenes and producing higher quality depth maps compared to handcrafted methods.
We propose CHOSEN, a simple yet flexible, robust and effective multi-view depth refinement framework. It can be employed in any existing multi-view stereo pipeline, with straightforward generalization capability for different multi-view capture systems such as camera relative positioning and lenses. Given an initial depth estimation, CHOSEN iteratively re-samples and selects the best hypotheses, and automatically adapts to different metric or intrinsic scales determined by the capture system. The key to our approach is the application of contrastive learning in an appropriate solution space and a carefully designed hypothesis feature, based on which positive and negative hypotheses can be effectively distinguished. Integrated in a simple baseline multi-view stereo pipeline, CHOSEN delivers impressive quality in terms of depth and normal accuracy compared to many current deep learning based multi-view stereo pipelines.
Pixel-wise regression tasks (e.g., monocular depth estimation (MDE) and optical flow estimation (OFE)) have been widely involved in our daily life in applications like autonomous driving, augmented reality and video composition. Although certain applications are security-critical or bear societal significance, the adversarial robustness of such models are not sufficiently studied, especially in the black-box scenario. In this work, we introduce the first unified black-box adversarial patch attack framework against pixel-wise regression tasks, aiming to identify the vulnerabilities of these models under query-based black-box attacks. We propose a novel square-based adversarial patch optimization framework and employ probabilistic square sampling and score-based gradient estimation techniques to generate the patch effectively and efficiently, overcoming the scalability problem of previous black-box patch attacks. Our attack prototype, named BadPart, is evaluated on both MDE and OFE tasks, utilizing a total of 7 models. BadPart surpasses 3 baseline methods in terms of both attack performance and efficiency. We also apply BadPart on the Google online service for portrait depth estimation, causing 43.5% relative distance error with 50K queries. State-of-the-art (SOTA) countermeasures cannot defend our attack effectively.
Semantic segmentation in bird's eye view (BEV) plays a crucial role in autonomous driving. Previous methods usually follow an end-to-end pipeline, directly predicting the BEV segmentation map from monocular RGB inputs. However, the challenge arises when the RGB inputs and BEV targets from distinct perspectives, making the direct point-to-point predicting hard to optimize. In this paper, we decompose the original BEV segmentation task into two stages, namely BEV map reconstruction and RGB-BEV feature alignment. In the first stage, we train a BEV autoencoder to reconstruct the BEV segmentation maps given corrupted noisy latent representation, which urges the decoder to learn fundamental knowledge of typical BEV patterns. The second stage involves mapping RGB input images into the BEV latent space of the first stage, directly optimizing the correlations between the two views at the feature level. Our approach simplifies the complexity of combining perception and generation into distinct steps, equipping the model to handle intricate and challenging scenes effectively. Besides, we propose to transform the BEV segmentation map from the Cartesian to the polar coordinate system to establish the column-wise correspondence between RGB images and BEV maps. Moreover, our method requires neither multi-scale features nor camera intrinsic parameters for depth estimation and saves computational overhead. Extensive experiments on nuScenes and Argoverse show the effectiveness and efficiency of our method. Code is available at https://github.com/happytianhao/TaDe.
In recent years, there have been significant advancements in 3D reconstruction and dense RGB-D SLAM systems. One notable development is the application of Neural Radiance Fields (NeRF) in these systems, which utilizes implicit neural representation to encode 3D scenes. This extension of NeRF to SLAM has shown promising results. However, the depth images obtained from consumer-grade RGB-D sensors are often sparse and noisy, which poses significant challenges for 3D reconstruction and affects the accuracy of the representation of the scene geometry. Moreover, the original hierarchical feature grid with occupancy value is inaccurate for scene geometry representation. Furthermore, the existing methods select random pixels for camera tracking, which leads to inaccurate localization and is not robust in real-world indoor environments. To this end, we present NeSLAM, an advanced framework that achieves accurate and dense depth estimation, robust camera tracking, and realistic synthesis of novel views. First, a depth completion and denoising network is designed to provide dense geometry prior and guide the neural implicit representation optimization. Second, the occupancy scene representation is replaced with Signed Distance Field (SDF) hierarchical scene representation for high-quality reconstruction and view synthesis. Furthermore, we also propose a NeRF-based self-supervised feature tracking algorithm for robust real-time tracking. Experiments on various indoor datasets demonstrate the effectiveness and accuracy of the system in reconstruction, tracking quality, and novel view synthesis.
Accurate monocular metric depth estimation (MMDE) is crucial to solving downstream tasks in 3D perception and modeling. However, the remarkable accuracy of recent MMDE methods is confined to their training domains. These methods fail to generalize to unseen domains even in the presence of moderate domain gaps, which hinders their practical applicability. We propose a new model, UniDepth, capable of reconstructing metric 3D scenes from solely single images across domains. Departing from the existing MMDE methods, UniDepth directly predicts metric 3D points from the input image at inference time without any additional information, striving for a universal and flexible MMDE solution. In particular, UniDepth implements a self-promptable camera module predicting dense camera representation to condition depth features. Our model exploits a pseudo-spherical output representation, which disentangles camera and depth representations. In addition, we propose a geometric invariance loss that promotes the invariance of camera-prompted depth features. Thorough evaluations on ten datasets in a zero-shot regime consistently demonstrate the superior performance of UniDepth, even when compared with methods directly trained on the testing domains. Code and models are available at: https://github.com/lpiccinelli-eth/unidepth
Acquiring accurate depth information of transparent objects using off-the-shelf RGB-D cameras is a well-known challenge in Computer Vision and Robotics. Depth estimation/completion methods are typically employed and trained on datasets with quality depth labels acquired from either simulation, additional sensors or specialized data collection setups and known 3d models. However, acquiring reliable depth information for datasets at scale is not straightforward, limiting training scalability and generalization. Neural Radiance Fields (NeRFs) are learning-free approaches and have demonstrated wide success in novel view synthesis and shape recovery. However, heuristics and controlled environments (lights, backgrounds, etc) are often required to accurately capture specular surfaces. In this paper, we propose using Visual Foundation Models (VFMs) for segmentation in a zero-shot, label-free way to guide the NeRF reconstruction process for these objects via the simultaneous reconstruction of semantic fields and extensions to increase robustness. Our proposed method Segmentation-AIDed NeRF (SAID-NeRF) shows significant performance on depth completion datasets for transparent objects and robotic grasping.
Self-supervised multi-frame methods have currently achieved promising results in depth estimation. However, these methods often suffer from mismatch problems due to the moving objects, which break the static assumption. Additionally, unfairness can occur when calculating photometric errors in high-freq or low-texture regions of the images. To address these issues, existing approaches use additional semantic priori black-box networks to separate moving objects and improve the model only at the loss level. Therefore, we propose FlowDepth, where a Dynamic Motion Flow Module (DMFM) decouples the optical flow by a mechanism-based approach and warps the dynamic regions thus solving the mismatch problem. For the unfairness of photometric errors caused by high-freq and low-texture regions, we use Depth-Cue-Aware Blur (DCABlur) and Cost-Volume sparsity loss respectively at the input and the loss level to solve the problem. Experimental results on the KITTI and Cityscapes datasets show that our method outperforms the state-of-the-art methods.
Self-supervised monocular depth estimation methods have been increasingly given much attention due to the benefit of not requiring large, labelled datasets. Such self-supervised methods require high-quality salient features and consequently suffer from severe performance drop for indoor scenes, where low-textured regions dominant in the scenes are almost indiscriminative. To address the issue, we propose a self-supervised indoor monocular depth estimation framework called $\mathrm{F^2Depth}$. A self-supervised optical flow estimation network is introduced to supervise depth learning. To improve optical flow estimation performance in low-textured areas, only some patches of points with more discriminative features are adopted for finetuning based on our well-designed patch-based photometric loss. The finetuned optical flow estimation network generates high-accuracy optical flow as a supervisory signal for depth estimation. Correspondingly, an optical flow consistency loss is designed. Multi-scale feature maps produced by finetuned optical flow estimation network perform warping to compute feature map synthesis loss as another supervisory signal for depth learning. Experimental results on the NYU Depth V2 dataset demonstrate the effectiveness of the framework and our proposed losses. To evaluate the generalization ability of our $\mathrm{F^2Depth}$, we collect a Campus Indoor depth dataset composed of approximately 1500 points selected from 99 images in 18 scenes. Zero-shot generalization experiments on 7-Scenes dataset and Campus Indoor achieve $\delta_1$ accuracy of 75.8% and 76.0% respectively. The accuracy results show that our model can generalize well to monocular images captured in unknown indoor scenes.