Current, self-supervised depth estimation architectures rely on clear and sunny weather scenes to train deep neural networks. However, in many locations, this assumption is too strong. For example in the UK (2021), 149 days consisted of rain. For these architectures to be effective in real-world applications, we must create models that can generalise to all weather conditions, times of the day and image qualities. Using a combination of computer graphics and generative models, one can augment existing sunny-weather data in a variety of ways that simulate adverse weather effects. While it is tempting to use such data augmentations for self-supervised depth, in the past this was shown to degrade performance instead of improving it. In this paper, we put forward a method that uses augmentations to remedy this problem. By exploiting the correspondence between unaugmented and augmented data we introduce a pseudo-supervised loss for both depth and pose estimation. This brings back some of the benefits of supervised learning while still not requiring any labels. We also make a series of practical recommendations which collectively offer a reliable, efficient framework for weather-related augmentation of self-supervised depth from monocular video. We present extensive testing to show that our method, Robust-Depth, achieves SotA performance on the KITTI dataset while significantly surpassing SotA on challenging, adverse condition data such as DrivingStereo, Foggy CityScape and NuScenes-Night. The project website can be found here https://kieran514.github.io/Robust-Depth-Project/.
We introduce the notion of point affiliation into feature upsampling. By abstracting a feature map into non-overlapped semantic clusters formed by points of identical semantic meaning, feature upsampling can be viewed as point affiliation -- designating a semantic cluster for each upsampled point. In the framework of kernel-based dynamic upsampling, we show that an upsampled point can resort to its low-res decoder neighbors and high-res encoder point to reason the affiliation, conditioned on the mutual similarity between them. We therefore present a generic formulation for generating similarity-aware upsampling kernels and prove that such kernels encourage not only semantic smoothness but also boundary sharpness. This formulation constitutes a novel, lightweight, and universal upsampling solution, Similarity-Aware Point Affiliation (SAPA). We show its working mechanism via our preliminary designs with window-shape kernel. After probing the limitations of the designs on object detection, we reveal additional insights for upsampling, leading to SAPA with the dynamic kernel shape. Extensive experiments demonstrate that SAPA outperforms prior upsamplers and invites consistent performance improvements on a number of dense prediction tasks, including semantic segmentation, object detection, instance segmentation, panoptic segmentation, image matting, and depth estimation. Code is made available at: https://github.com/tiny-smart/sapa
Recent work in unsupervised multi-object segmentation shows impressive results by predicting motion from a single image despite the inherent ambiguity in predicting motion without the next image. On the other hand, the set of possible motions for an image can be constrained to a low-dimensional space by considering the scene structure and moving objects in it. We propose to model pixel-wise geometry and object motion to remove ambiguity in reconstructing flow from a single image. Specifically, we divide the image into coherently moving regions and use depth to construct flow bases that best explain the observed flow in each region. We achieve state-of-the-art results in unsupervised multi-object segmentation on synthetic and real-world datasets by modeling the scene structure and object motion. Our evaluation of the predicted depth maps shows reliable performance in monocular depth estimation.
Video depth estimation aims to infer temporally consistent depth. One approach is to finetune a single-image model on each video with geometry constraints, which proves inefficient and lacks robustness. An alternative is learning to enforce consistency from data, which requires well-designed models and sufficient video depth data. To address both challenges, we introduce NVDS+ that stabilizes inconsistent depth estimated by various single-image models in a plug-and-play manner. We also elaborate a large-scale Video Depth in the Wild (VDW) dataset, which contains 14,203 videos with over two million frames, making it the largest natural-scene video depth dataset. Additionally, a bidirectional inference strategy is designed to improve consistency by adaptively fusing forward and backward predictions. We instantiate a model family ranging from small to large scales for different applications. The method is evaluated on VDW dataset and three public benchmarks. To further prove the versatility, we extend NVDS+ to video semantic segmentation and several downstream applications like bokeh rendering, novel view synthesis, and 3D reconstruction. Experimental results show that our method achieves significant improvements in consistency, accuracy, and efficiency. Our work serves as a solid baseline and data foundation for learning-based video depth estimation. Code and dataset are available at: https://github.com/RaymondWang987/NVDS
As demand for robotics manipulation application increases, accurate vision-based 6D pose estimation becomes essential for autonomous operations. Convolutional Neural Networks (CNNs) based approaches for pose estimation have been previously introduced. However, the quest for better performance still persists especially for accurate robotics manipulation. This quest extends to the Agri-robotics domain. In this paper, we propose TransPose, an improved Transformer-based 6D pose estimation with a depth refinement module. The architecture takes in only an RGB image as input with no additional supplementing modalities such as depth or thermal images. The architecture encompasses an innovative lighter depth estimation network that estimates depth from an RGB image using feature pyramid with an up-sampling method. A transformer-based detection network with additional prediction heads is proposed to directly regress the object's centre and predict the 6D pose of the target. A novel depth refinement module is then used alongside the predicted centers, 6D poses and depth patches to refine the accuracy of the estimated 6D pose. We extensively compared our results with other state-of-the-art methods and analysed our results for fruit-picking applications. The results we achieved show that our proposed technique outperforms the other methods available in the literature.
Despite the increasing prevalence of rotating-style capture (e.g., surveillance cameras), conventional stereo rectification techniques frequently fail due to the rotation-dominant motion and small baseline between views. In this paper, we tackle the challenge of performing stereo rectification for uncalibrated rotating cameras. To that end, we propose Depth-from-Rotation (DfR), a novel image rectification solution that analytically rectifies two images with two-point correspondences and serves for further depth estimation. Specifically, we model the motion of a rotating camera as the camera rotates on a sphere with fixed latitude. The camera's optical axis lies perpendicular to the sphere's surface. We call this latitudinal motion assumption. Then we derive a 2-point analytical solver from directly computing the rectified transformations on the two images. We also present a self-adaptive strategy to reduce the geometric distortion after rectification. Extensive synthetic and real data experiments demonstrate that the proposed method outperforms existing works in effectiveness and efficiency by a significant margin.
Stereo vision systems have become popular in computer vision applications, such as 3D reconstruction, object tracking, and autonomous navigation. However, traditional stereo vision systems that use rectilinear lenses may not be suitable for certain scenarios due to their limited field of view. This has led to the popularity of vision systems based on one or multiple fisheye cameras in different orientations, which can provide a field of view of 180x180 degrees or more. However, fisheye cameras introduce significant distortion at the edges that affects the accuracy of stereo matching and depth estimation. To overcome these limitations, this paper proposes a method for distortion-removal and depth estimation analysis for stereovision system using orthogonally divergent fisheye cameras (ODFC). The proposed method uses two virtual pinhole cameras (VPC), each VPC captures a small portion of the original view and presents it without any lens distortions, emulating the behavior of a pinhole camera. By carefully selecting the captured regions, it is possible to create a stereo pair using two VPCs. The performance of the proposed method is evaluated in both simulation using virtual environment and experiments using real cameras and their results compared to stereo cameras with parallel optical axes. The results demonstrate the effectiveness of the proposed method in terms of distortion removal and depth estimation accuracy.
One of the key problems in 3D object detection is to reduce the accuracy gap between methods based on LiDAR sensors and those based on monocular cameras. A recently proposed framework for monocular 3D detection based on Pseudo-Stereo has received considerable attention in the community. However, so far these two problems are discovered in existing practices, including (1) monocular depth estimation and Pseudo-Stereo detector must be trained separately, (2) Difficult to be compatible with different stereo detectors and (3) the overall calculation is large, which affects the reasoning speed. In this work, we propose an end-to-end, efficient pseudo-stereo 3D detection framework by introducing a Single-View Diffusion Model (SVDM) that uses a few iterations to gradually deliver right informative pixels to the left image. SVDM allows the entire pseudo-stereo 3D detection pipeline to be trained end-to-end and can benefit from the training of stereo detectors. Afterwards, we further explore the application of SVDM in depth-free stereo 3D detection, and the final framework is compatible with most stereo detectors. Among multiple benchmarks on the KITTI dataset, we achieve new state-of-the-art performance.
Neural Radiance Field (NeRF) has been a mainstream in novel view synthesis with its remarkable quality of rendered images and simple architecture. Although NeRF has been developed in various directions improving continuously its performance, the necessity of a dense set of multi-view images still exists as a stumbling block to progress for practical application. In this work, we propose FlipNeRF, a novel regularization method for few-shot novel view synthesis by utilizing our proposed flipped reflection rays. The flipped reflection rays are explicitly derived from the input ray directions and estimated normal vectors, and play a role of effective additional training rays while enabling to estimate more accurate surface normals and learn the 3D geometry effectively. Since the surface normal and the scene depth are both derived from the estimated densities along a ray, the accurate surface normal leads to more exact depth estimation, which is a key factor for few-shot novel view synthesis. Furthermore, with our proposed Uncertainty-aware Emptiness Loss and Bottleneck Feature Consistency Loss, FlipNeRF is able to estimate more reliable outputs with reducing floating artifacts effectively across the different scene structures, and enhance the feature-level consistency between the pair of the rays cast toward the photo-consistent pixels without any additional feature extractor, respectively. Our FlipNeRF achieves the SOTA performance on the multiple benchmarks across all the scenarios.
Monocular depth estimation is scale-ambiguous, and thus requires scale supervision to produce metric predictions. Even so, the resulting models will be geometry-specific, with learned scales that cannot be directly transferred across domains. Because of that, recent works focus instead on relative depth, eschewing scale in favor of improved up-to-scale zero-shot transfer. In this work we introduce ZeroDepth, a novel monocular depth estimation framework capable of predicting metric scale for arbitrary test images from different domains and camera parameters. This is achieved by (i) the use of input-level geometric embeddings that enable the network to learn a scale prior over objects; and (ii) decoupling the encoder and decoder stages, via a variational latent representation that is conditioned on single frame information. We evaluated ZeroDepth targeting both outdoor (KITTI, DDAD, nuScenes) and indoor (NYUv2) benchmarks, and achieved a new state-of-the-art in both settings using the same pre-trained model, outperforming methods that train on in-domain data and require test-time scaling to produce metric estimates.
As an emerging technology and a relatively affordable device, the 4D imaging radar has already been confirmed effective in performing 3D object detection in autonomous driving. Nevertheless, the sparsity and noisiness of 4D radar point clouds hinder further performance improvement, and in-depth studies about its fusion with other modalities are lacking. On the other hand, as a new image view transformation strategy, "sampling" has been applied in a few image-based detectors and shown to outperform the widely applied "depth-based splatting" proposed in Lift-Splat-Shoot (LSS), even without image depth prediction. However, the potential of "sampling" is not fully unleashed. This paper investigates the "sampling" view transformation strategy on the camera and 4D imaging radar fusion-based 3D object detection. LiDAR Excluded Lean (LXL) model, predicted image depth distribution maps and radar 3D occupancy grids are generated from image perspective view (PV) features and radar bird's eye view (BEV) features, respectively. They are sent to the core of LXL, called "radar occupancy-assisted depth-based sampling", to aid image view transformation. We demonstrated that more accurate view transformation can be performed by introducing image depths and radar information to enhance the "sampling" strategy. Experiments on VoD and TJ4DRadSet datasets show that the proposed method outperforms the state-of-the-art 3D object detection methods by a significant margin without bells and whistles. Ablation studies demonstrate that our method performs the best among different enhancement settings.
Depth perception is a crucial component of monoc-ular 3D detection tasks that typically involve ill-posed problems. In light of the success of sample mining techniques in 2D object detection, we propose a simple yet effective mining strategy for improving depth perception in 3D object detection. Concretely, we introduce a plain metric to evaluate the quality of depth predictions, which chooses the mined sample for the model. Moreover, we propose a Gradient-aware and Model-perceive Mining strategy (GMM) for depth learning, which exploits the predicted depth quality for better depth learning through easy mining. GMM is a general strategy that can be readily applied to several state-of-the-art monocular 3D detectors, improving the accuracy of depth prediction. Extensive experiments on the nuScenes dataset demonstrate that the proposed methods significantly improve the performance of 3D object detection while outperforming other state-of-the-art sample mining techniques by a considerable margin. On the nuScenes benchmark, GMM achieved the state-of-the-art (42.1% mAP and 47.3% NDS) performance in monocular object detection.
Dense pixel-specific representation learning at scale has been bottlenecked due to the unavailability of large-scale multi-view datasets. Current methods for building effective pretraining datasets heavily rely on annotated 3D meshes, point clouds, and camera parameters from simulated environments, preventing them from building datasets from real-world data sources where such metadata is lacking. We propose a pretraining dataset-curation approach that does not require any additional annotations. Our method allows us to generate multi-view datasets from both real-world videos and simulated environments at scale. Specifically, we experiment with two scales: MIMIC-1M with 1.3M and MIMIC-3M with 3.1M multi-view image pairs. We train multiple models with different masked image modeling objectives to showcase the following findings: Representations trained on our automatically generated MIMIC-3M outperform those learned from expensive crowdsourced datasets (ImageNet-1K) and those learned from synthetic environments (MULTIVIEW-HABITAT) on two dense geometric tasks: depth estimation on NYUv2 (1.7%), and surface normals estimation on Taskonomy (2.05%). For dense tasks which also require object understanding, we outperform MULTIVIEW-HABITAT, on semantic segmentation on ADE20K (3.89%), pose estimation on MSCOCO (9.4%), and reduce the gap with models pre-trained on the object-centric expensive ImageNet-1K. We outperform even when the representations are frozen, and when downstream training data is limited to few-shot. Larger dataset (MIMIC-3M) significantly improves performance, which is promising since our curation method can arbitrarily scale to produce even larger datasets. MIMIC code, dataset, and pretrained models are open-sourced at https://github.com/RAIVNLab/MIMIC.
Depth completion is the task of recovering dense depth maps from sparse ones, usually with the help of color images. Existing image-guided methods perform well on daytime depth perception self-driving benchmarks, but struggle in nighttime scenarios with poor visibility and complex illumination. To address these challenges, we propose a simple yet effective framework called LDCNet. Our key idea is to use Recurrent Inter-Convolution Differencing (RICD) and Illumination-Affinitive Intra-Convolution Differencing (IAICD) to enhance the nighttime color images and reduce the negative effects of the varying illumination, respectively. RICD explicitly estimates global illumination by differencing two convolutions with different kernels, treating the small-kernel-convolution feature as the center of the large-kernel-convolution feature in a new perspective. IAICD softly alleviates local relative light intensity by differencing a single convolution, where the center is dynamically aggregated based on neighboring pixels and the estimated illumination map in RICD. On both nighttime depth completion and depth estimation tasks, extensive experiments demonstrate the effectiveness of our LDCNet, reaching the state of the art.
Structured light has proven instrumental in 3D imaging, LiDAR, and holographic light projection. Metasurfaces, comprised of sub-wavelength-sized nanostructures, facilitate 180$^\circ$ field-of-view (FoV) structured light, circumventing the restricted FoV inherent in traditional optics like diffractive optical elements. However, extant metasurface-facilitated structured light exhibits sub-optimal performance in downstream tasks, due to heuristic pattern designs such as periodic dots that do not consider the objectives of the end application. In this paper, we present neural 360$^\circ$ structured light, driven by learned metasurfaces. We propose a differentiable framework, that encompasses a computationally-efficient 180$^\circ$ wave propagation model and a task-specific reconstructor, and exploits both transmission and reflection channels of the metasurface. Leveraging a first-order optimizer within our differentiable framework, we optimize the metasurface design, thereby realizing neural 360$^\circ$ structured light. We have utilized neural 360$^\circ$ structured light for holographic light projection and 3D imaging. Specifically, we demonstrate the first 360$^\circ$ light projection of complex patterns, enabled by our propagation model that can be computationally evaluated 50,000$\times$ faster than the Rayleigh-Sommerfeld propagation. For 3D imaging, we improve depth-estimation accuracy by 5.09$\times$ in RMSE compared to the heuristically-designed structured light. Neural 360$^\circ$ structured light promises robust 360$^\circ$ imaging and display for robotics, extended-reality systems, and human-computer interactions.
Automated driving systems use multi-modal sensor suites to ensure the reliable, redundant and robust perception of the operating domain, for example camera and LiDAR. An accurate extrinsic calibration is required to fuse the camera and LiDAR data into a common spatial reference frame required by high-level perception functions. Over the life of the vehicle the value of the extrinsic calibration can change due physical disturbances, introducing an error into the high-level perception functions. Therefore there is a need for continuous online extrinsic calibration algorithms which can automatically update the value of the camera-LiDAR calibration during the life of the vehicle using only sensor data. We propose using mutual information between the camera image's depth estimate, provided by commonly available monocular depth estimation networks, and the LiDAR pointcloud's geometric distance as a optimization metric for extrinsic calibration. Our method requires no calibration target, no ground truth training data and no expensive offline optimization. We demonstrate our algorithm's accuracy, precision, speed and self-diagnosis capability on the KITTI-360 data set.
Numerous studies have investigated the pivotal role of reliable 3D volume representation in scene perception tasks, such as multi-view stereo (MVS) and semantic scene completion (SSC). They typically construct 3D probability volumes directly with geometric correspondence, attempting to fully address the scene perception tasks in a single forward pass. However, such a single-step solution makes it hard to learn accurate and convincing volumetric probability, especially in challenging regions like unexpected occlusions and complicated light reflections. Therefore, this paper proposes to decompose the complicated 3D volume representation learning into a sequence of generative steps to facilitate fine and reliable scene perception. Considering the recent advances achieved by strong generative diffusion models, we introduce a multi-step learning framework, dubbed as VPD, dedicated to progressively refining the Volumetric Probability in a Diffusion process. Extensive experiments are conducted on scene perception tasks including multi-view stereo (MVS) and semantic scene completion (SSC), to validate the efficacy of our method in learning reliable volumetric representations. Notably, for the SSC task, our work stands out as the first to surpass LiDAR-based methods on the SemanticKITTI dataset.
Cutting-edge connected vehicle (CV) technologies have drawn much attention in recent years. The real-time traffic data captured by a CV can be shared with other CVs and data centers so as to open new possibilities for solving diverse transportation problems. However, imagery captured by onboard cameras in a connected environment, are not sufficiently investigated, especially for safety and health-oriented visual perception. In this paper, a bidirectional process of image synthesis and decomposition (BPISD) approach is proposed, and thus a novel self-supervised multi-task learning framework, to simultaneously estimate depth map, atmospheric visibility, airlight, and PM2.5 mass concentration, in which depth map and visibility are considered highly associated with traffic safety, while airlight and PM2.5 mass concentration are directly correlated with human health. Both the training and testing phases of the proposed system solely require a single image as input. Due to the innovative training pipeline, the depth estimation network can manage various levels of visibility conditions and overcome inherent problems in current image-synthesis-based depth estimation, thereby generating high-quality depth maps even in low-visibility situations and further benefiting accurate estimations of visibility, airlight, and PM2.5 mass concentration. Extensive experiments on the synthesized data from the KITTI and real-world data collected in Beijing demonstrate that the proposed method can (1) achieve performance competitive in depth estimation as compared with state-of-the-art methods when taking clear images as input; (2) predict vivid depth map for images contaminated by various levels of haze; and (3) accurately estimate visibility, airlight, and PM2.5 mass concentrations. Beneficial applications can be developed based on the presented work to improve traffic safety, air quality, and public health.
Monocular 3D object detection aims to locate objects in different scenes with just a single image. Due to the absence of depth information, several monocular 3D detection techniques have emerged that rely on auxiliary depth maps from the depth estimation task. There are multiple approaches to understanding the representation of depth maps, including treating them as pseudo-LiDAR point clouds, leveraging implicit end-to-end learning of depth information, or considering them as an image input. However, these methods have certain drawbacks, such as their reliance on the accuracy of estimated depth maps and suboptimal utilization of depth maps due to their image-based nature. While LiDAR-based methods and convolutional neural networks (CNNs) can be utilized for pseudo point clouds and depth maps, respectively, it is always an alternative. In this paper, we propose a framework named the Adaptive Distance Interval Separation Network (ADISN) that adopts a novel perspective on understanding depth maps, as a form that lies between LiDAR and images. We utilize an adaptive separation approach that partitions the depth map into various subgraphs based on distance and treats each of these subgraphs as an individual image for feature extraction. After adaptive separations, each subgraph solely contains pixels within a learned interval range. If there is a truncated object within this range, an evident curved edge will appear, which we can leverage for texture extraction using CNNs to obtain rich depth information in pixels. Meanwhile, to mitigate the inaccuracy of depth estimation, we designed an uncertainty module. To take advantage of both images and depth maps, we use different branches to learn localization detection tasks and appearance tasks separately.
Depth estimation is a cornerstone of perception in autonomous driving and robotic systems. The considerable cost and relatively sparse data acquisition of LiDAR systems have led to the exploration of cost-effective alternatives, notably, self-supervised depth estimation. Nevertheless, current self-supervised depth estimation methods grapple with several limitations: (1) the failure to adequately leverage informative multi-camera views. (2) the limited capacity to handle dynamic objects effectively. To address these challenges, we present BEVScope, an innovative approach to self-supervised depth estimation that harnesses Bird's-Eye-View (BEV) features. Concurrently, we propose an adaptive loss function, specifically designed to mitigate the complexities associated with moving objects. Empirical evaluations conducted on the Nuscenes dataset validate our approach, demonstrating competitive performance. Code will be released at https://github.com/myc634/BEVScope.