skip to main content

Title: Synthesizing 3D Shapes From Silhouette Image Collections Using Multi-Projection Generative Adversarial Networks
We present a new weakly supervised learning-based method for generating novel category-specific 3D shapes from unoccluded image collections. Our method is weakly supervised and only requires silhouette annotations from unoccluded, category-specific objects. Our method does not require access to the object's 3D shape, multiple observations per object from different views, intra-image pixel correspondences, or any view annotations. Key to our method is a novel multi-projection generative adversarial network (MP-GAN) that trains a 3D shape generator to be consistent with multiple 2D projections of the 3D shapes, and without direct access to these 3D shapes. This is achieved through multiple discriminators that encode the distribution of 2D projections of the 3D shapes seen from a different views. Additionally, to determine the view information for each silhouette image, we also train a view prediction network on visualizations of 3D shapes synthesized by the generator. We iteratively alternate between training the generator and training the view prediction network. We validate our multi-projection GAN on both synthetic and real image datasets. Furthermore, we also show that multi-projection GANs can aid in learning other high-dimensional distributions from lower dimensional training datasets, such as material-class specific spatially varying reflectance properties from images.
Authors:
; ; ;
Award ID(s):
1909028 1350323
Publication Date:
NSF-PAR ID:
10167638
Journal Name:
2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Page Range or eLocation-ID:
5530 to 5539
Sponsoring Org:
National Science Foundation
More Like this
  1. This paper introduces a deep neural network based method, i.e., DeepOrganNet, to generate and visualize fully high-fidelity 3D / 4D organ geometric models from single-view medical images with complicated background in real time. Traditional 3D / 4D medical image reconstruction requires near hundreds of projections, which cost insufferable computational time and deliver undesirable high imaging / radiation dose to human subjects. Moreover, it always needs further notorious processes to segment or extract the accurate 3D organ models subsequently. The computational time and imaging dose can be reduced by decreasing the number of projections, but the reconstructed image quality is degradedmore »accordingly. To our knowledge, there is no method directly and explicitly reconstructing multiple 3D organ meshes from a single 2D medical grayscale image on the fly. Given single-view 2D medical images, e.g., 3D / 4D-CT projections or X-ray images, our end-to-end DeepOrganNet framework can efficiently and effectively reconstruct 3D / 4D lung models with a variety of geometric shapes by learning the smooth deformation fields from multiple templates based on a trivariate tensor-product deformation technique, leveraging an informative latent descriptor extracted from input 2D images. The proposed method can guarantee to generate high-quality and high-fidelity manifold meshes for 3D / 4D lung models; while, all current deep learning based approaches on the shape reconstruction from a single image cannot. The major contributions of this work are to accurately reconstruct the 3D organ shapes from 2D single-view projection, significantly improve the procedure time to allow on-the-fly visualization, and dramatically reduce the imaging dose for human subjects. Experimental results are evaluated and compared with the traditional reconstruction method and the state-of-the-art in deep learning, by using extensive 3D and 4D examples, including both synthetic phantom and real patient datasets. The efficiency of the proposed method shows that it only needs several milliseconds to generate organ meshes with 10K vertices, which has great potential to be used in real-time image guided radiation therapy (IGRT).« less
  2. 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 ismore »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.« less
  3. The success of supervised learning requires large-scale ground truth labels which are very expensive, time- consuming, or may need special skills to annotate. To address this issue, many self- or un-supervised methods are developed. Unlike most existing self-supervised methods to learn only 2D image features or only 3D point cloud features, this paper presents a novel and effective self-supervised learning approach to jointly learn both 2D image features and 3D point cloud features by exploiting cross-modality and cross-view correspondences without using any human annotated labels. Specifically, 2D image features of rendered images from different views are extracted by a 2Dmore »convolutional neural network, and 3D point cloud features are extracted by a graph convolution neural network. Two types of features are fed into a two-layer fully connected neural network to estimate the cross-modality correspondence. The three networks are jointly trained (i.e. cross-modality) by verifying whether two sampled data of different modalities belong to the same object, meanwhile, the 2D convolutional neural network is additionally optimized through minimizing intra-object distance while maximizing inter-object distance of rendered images in different views (i.e. cross-view). The effectiveness of the learned 2D and 3D features is evaluated by transferring them on five different tasks including multi-view 2D shape recognition, 3D shape recognition, multi-view 2D shape retrieval, 3D shape retrieval, and 3D part-segmentation. Extensive evaluations on all the five different tasks across different datasets demonstrate strong generalization and effectiveness of the learned 2D and 3D features by the proposed self-supervised method.« less
  4. This paper presents a semi-supervised learning framework for a customized semantic segmentation task using multiview image streams. A key challenge of the customized task lies in the limited accessibility of the labeled data due to the requirement of prohibitive manual annotation effort. We hypothesize that it is possible to leverage multiview image streams that are linked through the underlying 3D geometry, which can provide an additional supervisionary signal to train a segmentation model. We formulate a new cross-supervision method using a shape belief transfer---the segmentation belief in one image is used to predict that of the other image through epipolarmore »geometry analogous to shape-from-silhouette. The shape belief transfer provides the upper and lower bounds of the segmentation for the unlabeled data where its gap approaches asymptotically to zero as the number of the labeled views increases. We integrate this theory to design a novel network that is agnostic to camera calibration, network model, and semantic category and bypasses the intermediate process of suboptimal 3D reconstruction. We validate this network by recognizing a customized semantic category per pixel from realworld visual data including non-human species and a subject of interest in social videos where attaining large-scale annotation data is infeasible.« less
  5. We investigate the problem of learning to generate 3D parametric surface representations for novel object instances, as seen from one or more views. Previous work on learning shape reconstruction from multiple views uses discrete representations such as point clouds or voxels, while continuous surface generation approaches lack multi-view consistency. We address these issues by designing neural networks capable of generating high-quality parametric 3D surfaces which are also consistent between views. Furthermore, the generated 3D surfaces preserve accurate image pixel to 3D surface point correspondences, allowing us to lift texture information to reconstruct shapes with rich geometry and appearance. Our methodmore »is supervised and trained on a public dataset of shapes from common object categories. Quantitative results indicate that our method significantly outperforms previous work, while qualitative results demonstrate the high quality of our reconstructions.« less