Contrastive Learning for 3D Point Clouds Classification and Shape Completion

In this paper, we present the idea of Self Supervised learning on the shape completion and classification of point clouds. Most 3D shape completion pipelines utilize AutoEncoders to extract features from point clouds used in downstream tasks such as classification, segmentation, detection, and other related applications. Our idea is to add contrastive learning into AutoEncoders to encourage global feature learning of the point cloud classes. It is performed by optimizing triplet loss. Furthermore, local feature representations learning of point cloud is performed by adding the Chamfer distance function. To evaluate the performance of our approach, we utilize the PointNet classifier. We also extend the number of classes for evaluation from 4 to 10 to show the generalization ability of the learned features. Based on our results, embeddings generated from the contrastive AutoEncoder enhances shape completion and classification performance from 84.2% to 84.9% of point clouds achieving the state-of-the-art results with 10 classes.

Contrasitive Learning for 3D Point Clouds Classification and Shape Completion

In this paper, we present the idea of Self Supervised learning on the Shape Completion and Classification of point clouds. Most 3D shape completion pipelines utilize autoencoders to extract features from point clouds used in downstream tasks such as Classification, Segmentation, Detection, and other related applications. Our idea is to add Contrastive Learning into Auto-Encoders to learn both global and local feature representations of point clouds. We use a combination of Triplet Loss and Chamfer distance to learn global and local feature representations. To evaluate the performance of embeddings for Classification, we utilize the PointNet classifier. We also extend the number of classes to evaluate our model from 4 to 10 to show the generalization ability of learned features. Based on our results, embedding generated from the Contrastive autoencoder enhances Shape Completion and Classification performance from 84.2% to 84.9% of point clouds achieving the state-of-the-art results with 10 classes.

HUMANNET-a Two-Tiered Deep Neural Network Architecture for Self-Occluding Humanoid Pose Reconstruction

Majority of current research focuses on a single static object reconstruction from a given pointcloud. However, the existing approaches are not applicable to real world applications such as dynamic and morphing scene reconstruction. To solve this, we propose a novel two-tiered deep neural network architecture, which is capable of reconstructing self-obstructed human-like morphing shapes from a depth frame in conjunction with cameras intrinsic parameters. The tests were performed using on custom dataset generated using a combination of AMASS and MoVi datasets. The proposed network achieved Jaccards’ Index of 0.7907 for the first tier, which is used to extract region of interest from the point cloud. The second tier of the network has achieved Earth Mover’s distance of 0.0256 and Chamfer distance of 0.276, indicating good experimental results. Further, subjective reconstruction results inspection shows strong predictive capabilities of the network, with the solution being able to reconstruct limb positions from very few object details.

Evaluation of four point cloud similarity measures for the use in autonomous driving

Measuring the similarity between point clouds is required in many areas. In autonomous driving, point clouds for 3D perception are estimated from camera images but these estimations are error-prone. Furthermore, there is a lack of measures for quality quantification using ground truth. In this paper, we derive conditions point cloud comparisons need to fulfill and accordingly evaluate the Chamfer distance, a lower bound of the Gromov Wasserstein metric, and the ratio measure. We show that the ratio measure is not affected by erroneous points and therefore introduce the new measure “average ratio”. All measures are evaluated and compared using exemplary point clouds. We discuss characteristics, advantages and drawbacks with respect to interpretability, noise resistance, environmental representation, and computation.

GENERATIVE NETWORKS FOR POINT CLOUD GENERATION IN CULTURAL HERITAGE DOMAIN

In the Cultural Heritage (CH) domain, the semantic segmentation of 3D point clouds with Deep Learning (DL) techniques allows to recognize historical architectural elements, at a suitable level of detail, and hence expedite the process of modelling historical buildings for the development of BIM models from survey data. However, it is more difficult to collect a balanced dataset of labelled architectural elements for training a network. In fact, the CH objects are unique, and it is challenging for the network to recognize this kind of data. In recent years, Generative Networks have proven to be proper for generating new data. Starting from such premises, in this paper Generative Networks have been used for augmenting a CH dataset. In particular, the performances of three state-of-art Generative Networks such as PointGrow, PointFLow and PointGMM have been compared in terms of Jensen-Shannon Divergence (JSD), the Minimum Matching Distance-Chamfer Distance (MMD-CD) and the Minimum Matching Distance-Earth Mover’s Distance (MMD-EMD). The objects generated have been used for augmenting two classes of ArCH dataset, which are columns and windows. Then a DGCNN-Mod network was trained and tested for the semantic segmentation task, comparing the performance in the case of the ArCH dataset without and with augmentation.

ANALYSIS OF ROTATION AND SCALE INVARIANCE FOR PROXIMITY MEASURES IN TARGET DETECTION

The problem of stability of object detection in images using proximity measures is considered. The purpose of the work is to determine the degree of invariance of various proximity measures for detecting objects by reference when rotating and zooming the scanned image. The proximity measure that is most resistant to these geometric transformations of the image is found out. The proximity measures are analyzed: correlation, comparison, Chamfer Distance. The target location is based on the coordinates of the extremum of the target function. Modeling is performed in the Matlab software package. A database of thirty television images was created to test the proximity measures. Test images contain the required objects and imitations of both complex and simple backgrounds. It was determined that all considered proximity measures steadily determine the target with small turns and scaling factors.

Morphing and Sampling Network for Dense Point Cloud Completion

3D point cloud completion, the task of inferring the complete geometric shape from a partial point cloud, has been attracting attention in the community. For acquiring high-fidelity dense point clouds and avoiding uneven distribution, blurred details, or structural loss of existing methods' results, we propose a novel approach to complete the partial point cloud in two stages. Specifically, in the first stage, the approach predicts a complete but coarse-grained point cloud with a collection of parametric surface elements. Then, in the second stage, it merges the coarse-grained prediction with the input point cloud by a novel sampling algorithm. Our method utilizes a joint loss function to guide the distribution of the points. Extensive experiments verify the effectiveness of our method and demonstrate that it outperforms the existing methods in both the Earth Mover's Distance (EMD) and the Chamfer Distance (CD).

Transductive Zero-Shot Learning via Visual Center Adaptation

In this paper, we propose a Visual Center Adaptation Method (VCAM) to address the domain shift problem in zero-shot learning. For the seen classes in the training data, VCAM builds an embedding space by learning the mapping from semantic space to some visual centers. While for unseen classes in the test data, the construction of embedding space is constrained by a symmetric Chamfer-distance term, aiming to adapt the distribution of the synthetic visual centers to that of the real cluster centers. Therefore the learned embedding space can generalize the unseen classes well. Experiments on two widely used datasets demonstrate that our model significantly outperforms state-of-the-art methods.