We present 3DVNet, a novel multi-view stereo (MVS) depth-prediction method that combines the advantages of previous depth-based and volumetric MVS approaches. Our key idea is the use of a 3D scene-modeling network that iteratively updates a set of coarse depth predictions, resulting in highly accurate predictions which agree on the underlying scene geometry. Unlike existing depth-prediction techniques, our method uses a volumetric 3D convolutional neural network (CNN) that operates in world space on all depth maps jointly. The network can therefore learn meaningful scene-level priors. Furthermore, unlike existing volumetric MVS techniques, our 3D CNN operates on a feature-augmented point cloud, allowing for effective aggregation of multi-view information and flexible iterative refinement of depth maps. Experimental results show our method exceeds state-of-the-art accuracy in both depth prediction and 3D reconstruction metrics on the ScanNet dataset, as well as a selection of scenes from the TUM-RGBD and ICL-NUIM datasets. This shows that our method is both effective and generalizes to new settings.
more »
« less
Towards Efficient 3D Point Cloud Scene Completion via Novel Depth View Synthesis
3D point cloud completion has been a long-standing challenge at scale, and corresponding per-point supervised training strategies suffered from cumbersome annotations. 2D supervision has recently emerged as a promising alternative for 3D tasks, but specific approaches for 3D point cloud completion still remain to be explored. To overcome these limitations, we propose an end-to-end method that directly lifts a single depth map to a completed point cloud. With one depth map as input, a multi-way novel depth view synthesis network (NDVNet) is designed to infer coarsely completed depth maps under various viewpoints. Meanwhile, a geometric depth perspective rendering module is introduced to utilize the raw input depth map to generate a reprojected depth map for each view. Therefore, the two parallelly generated depth maps for each view are further concatenated and refined by a depth completion network (DCNet). The final completed point cloud is fused from all refined depth views. Experimental results demonstrate the effectiveness of our proposed approach composed of aforementioned components, to produce high-quality, state-of-the-art results on the popular SUNCG benchmark.
more »
« less
- Award ID(s):
- 2041307
- NSF-PAR ID:
- 10279865
- Date Published:
- Journal Name:
- International Conference on Pattern Recognition (ICPR), 2020.
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
more » « less
Abstract Teeth scans are essential for many applications in orthodontics, where the teeth structures are virtualized to facilitate the design and fabrication of the prosthetic piece. Nevertheless, due to the limitations caused by factors such as viewing angles, occlusions, and sensor resolution, the 3D scanned point clouds (PCs) could be noisy or incomplete. Hence, there is a critical need to enhance the quality of the teeth PCs to ensure a suitable dental treatment. Toward this end, we propose a systematic framework including a two-step data augmentation (DA) technique to augment the limited teeth PCs and a hybrid deep learning (DL) method to complete the incomplete PCs. For the two-step DA, we first mirror and combine the PCs based on the bilateral symmetry of the human teeth and then augment the PCs based on an iterative generative adversarial network (GAN). Two filters are designed to avoid the outlier and duplicated PCs during the DA. For the hybrid DL, we first use a deep autoencoder (AE) to represent the PCs. Then, we propose a hybrid approach that selects the best completion to the teeth PCs from AE and a reinforcement learning (RL) agent-controlled GAN. Ablation study is performed to analyze each component’s contribution. We compared our method with other benchmark methods including point cloud network (PCN), cascaded refinement network (CRN), and variational relational point completion network (VRC-Net), and demonstrated that the proposed framework is suitable for completing teeth PCs with good accuracy over different scenarios.
-
null (Ed.)In this paper, we address the problem of estimating dense depth from a sequence of images using deep neural networks. Specifically, we employ a dense-optical-flow network to compute correspondences and then triangulate the point cloud to obtain an initial depth map. Parts of the point cloud, however, may be less accurate than others due to lack of common observations or small parallax. To further increase the triangulation accuracy, we introduce a depth-refinement network (DRN) that optimizes the initial depth map based on the image’s contextual cues. In particular, the DRN contains an iterative refinement module (IRM) that improves the depth accuracy over iterations by refining the deep features. Lastly, the DRN also predicts the uncertainty in the refined depths, which is desirable in applications such as measurement selection for scene reconstruction. We show experimentally that our algorithm outperforms state-of-the-art approaches in terms of depth accuracy, and verify that our predicted uncertainty is highly correlated to the actual depth error.more » « less
-
Background: Drivers gather most of the information they need to drive by looking at the world around them and at visual displays within the vehicle. Navigation systems automate the way drivers navigate. In using these systems, drivers offload both tactical (route following) and strategic aspects (route planning) of navigational tasks to the automated SatNav system, freeing up cognitive and attentional resources that can be used in other tasks (Burnett, 2009). Despite the potential benefits and opportunities that navigation systems provide, their use can also be problematic. For example, research suggests that drivers using SatNav do not develop as much environmental spatial knowledge as drivers using paper maps (Waters & Winter, 2011; Parush, Ahuvia, & Erev, 2007). With recent growth and advances of augmented reality (AR) head-up displays (HUDs), there are new opportunities to display navigation information directly within a driver’s forward field of view, allowing them to gather information needed to navigate without looking away from the road. While the technology is promising, the nuances of interface design and its impacts on drivers must be further understood before AR can be widely and safely incorporated into vehicles. Specifically, an impact that warrants investigation is the role of AR HUDS in spatial knowledge acquisition while driving. Acquiring high levels of spatial knowledge is crucial for navigation tasks because individuals who have greater levels of spatial knowledge acquisition are more capable of navigating based on their own internal knowledge (Bolton, Burnett, & Large, 2015). Moreover, the ability to develop an accurate and comprehensive cognitive map acts as a social function in which individuals are able to navigate for others, provide verbal directions and sketch direction maps (Hill, 1987). Given these points, the relationship between spatial knowledge acquisition and novel technologies such as AR HUDs in driving is a relevant topic for investigation. Objectives: This work explored whether providing conformal AR navigational cues improves spatial knowledge acquisition (as compared to traditional HUD visual cues) to assess the plausibility and justification for investment in generating larger FOV AR HUDs with potentially multiple focal planes. Methods: This study employed a 2x2 between-subjects design in which twenty-four participants were counterbalanced by gender. We used a fixed base, medium fidelity driving simulator for where participants drove while navigating with one of two possible HUD interface designs: a world-relative arrow post sign and a screen-relative traditional arrow. During the 10-15 minute drive, participants drove the route and were encouraged to verbally share feedback as they proceeded. After the drive, participants completed a NASA-TLX questionnaire to record their perceived workload. We measured spatial knowledge at two levels: landmark and route knowledge. Landmark knowledge was assessed using an iconic recognition task, while route knowledge was assessed using a scene ordering task. After completion of the study, individuals signed a post-trial consent form and were compensated $10 for their time. Results: NASA-TLX performance subscale ratings revealed that participants felt that they performed better during the world-relative condition but at a higher rate of perceived workload. However, in terms of perceived workload, results suggest there is no significant difference between interface design conditions. Landmark knowledge results suggest that the mean number of remembered scenes among both conditions is statistically similar, indicating participants using both interface designs remembered the same proportion of on-route scenes. Deviance analysis show that only maneuver direction had an influence on landmark knowledge testing performance. Route knowledge results suggest that the proportion of scenes on-route which were correctly sequenced by participants is similar under both conditions. Finally, participants exhibited poorer performance in the route knowledge task as compared to landmark knowledge task (independent of HUD interface design). Conclusions: This study described a driving simulator study which evaluated the head-up provision of two types of AR navigation interface designs. The world-relative condition placed an artificial post sign at the corner of an approaching intersection containing a real landmark. The screen-relative condition displayed turn directions using a screen-fixed traditional arrow located directly ahead of the participant on the right or left side on the HUD. Overall results of this initial study provide evidence that the use of both screen-relative and world-relative AR head-up display interfaces have similar impact on spatial knowledge acquisition and perceived workload while driving. These results contrast a common perspective in the AR community that conformal, world-relative graphics are inherently more effective. This study instead suggests that simple, screen-fixed designs may indeed be effective in certain contexts.more » « less
-
In this work, we tackle the problem of category-level online pose tracking of objects from point cloud sequences. For the first time, we propose a unified framework that can handle 9DoF pose tracking for novel rigid object instances as well as per-part pose tracking for articulated objects from known categories. Here the 9DoF pose, comprising 6D pose and 3D size, is equivalent to a 3D amodal bounding box representation with free 6D pose. Given the depth point cloud at the current frame and the estimated pose from the last frame, our novel end-to-end pipeline learns to accurately update the pose. Our pipeline is composed of three modules: 1) a pose canonicalization module that normalizes the pose of the input depth point cloud; 2) RotationNet, a module that directly regresses small interframe delta rotations; and 3) CoordinateNet, a module that predicts the normalized coordinates and segmentation, enabling analytical computation of the 3D size and translation. Leveraging the small pose regime in the pose-canonicalized point clouds, our method integrates the best of both worlds by combining dense coordinate prediction and direct rotation regression, thus yielding an end-to-end differentiable pipeline optimized for 9DoF pose accuracy (without using non-differentiable RANSAC). Our extensive experiments demonstrate that our method achieves new state-of-the-art performance on category-level rigid object pose (NOCSREAL275 [29]) and articulated object pose benchmarks (SAPIEN [34], BMVC [18]) at the fastest FPS ∼ 12.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
