Search

Recovering the 3D shape of an object from single or multiple images with deep neural networks has been attracting increasing attention in the past few years. Mainstream works (e.g. 3D-R2N2) use recurrent neural networks (RNNs) to sequentially fuse feature maps of input images. However, RNN-based approaches are unable to produce consistent reconstruction results when given the same input images with different orders. Moreover, RNNs may forget important features from early input images due to long-term memory loss. To address these issues, we propose a novel framework for single-view and multi-view 3D object reconstruction, named Pix2Vox++. By using a well-designed encoder-decoder, it generates a coarse 3D volume from each input image. A multi-scale context-aware fusion module is then introduced to adaptively select high-quality reconstructions for different parts from all coarse 3D volumes to obtain a fused 3D volume. To further correct the wrongly recovered parts in the fused 3D volume, a refiner is adopted to generate the final output. Experimental results on the ShapeNet, Pix3D, and Things3D benchmarks show that Pix2Vox++ performs favorably against state-of-the-art methods in terms of both accuracy and efficiency. [ABSTRACT FROM AUTHOR]

Graph convolution networks (GCNs) based methods for 3D human pose estimation usually aggregate immediate features of single-hop nodes, which are unaware of the correlation of multi-hop nodes and therefore neglect long-range dependency for predicting complex poses. In addition, they typically operate either on single-scale or sequential down-sampled multi-scale graph representations, resulting in the loss of contextual information or spatial details. To address these problems, this paper proposes a parallel hop-aware graph attention network (PHGANet) for 3D human pose estimation, which learns enriched hop-aware correlation of the skeleton joints while maintaining the spatially-precise representations of the human graph. Specifically, we propose a hop-aware skeletal graph attention (HSGAT) module to capture the semantic correlation of multi-hop nodes, which first calculates skeleton-based 1-hop attention and then disseminates it to arbitrary hops via graph connectivity. To alleviate the redundant noise introduced by the interactions with distant nodes, HSGAT uses an attenuation strategy to separate attention from distinct hops and assign them learnable attenuation weights according to their distances adaptively. Upon HSGAT, we further build PHGANet with multiple parallel branches of stacked HSGAT modules to learn the enriched hop-aware correlation of human skeletal structures at different scales. In addition, a joint centrality encoding scheme is proposed to introduce node importance as a bias in the learned graph representation, which makes the core joints (e.g., neck and pelvis) more influential during node aggregation. Experimental results indicate that PHGANet performs favorably against state-of-the-art methods on the Human3.6M and MPI-INF-3DHP benchmarks. Models and code are available at https://github.com/ChenyangWang95/PHGANet/. [ABSTRACT FROM AUTHOR]

With the popularization of mobile devices, recent years have witnessed an emerging potential for mobile landmark search. In this scenario, the user experience heavily depends on the efficiency of query transmission over a wireless link. As sending a query photo is time consuming, recent works have proposed to extract compact visual descriptors directly on the mobile end towards low bit rate transmission. Typically, these descriptors are extracted based solely on the visual content of a query, and the location cues from the mobile end are rarely exploited. In this paper, we present a Location Discriminative Vocabulary Coding (LDVC) scheme, which achieves extremely low bit rate query transmission, discriminative landmark description, as well as scalable descriptor delivery in a unified framework. Our first contribution is a compact and location discriminative visual landmark descriptor, which is offline learnt in two-step: First, we adopt spectral clustering to segment a city map into distinct geographical regions, where both visual and geographical similarities are fused to optimize the partition of city-scale geo-tagged photos. Second, we propose to learn LDVC in each region with two schemes: (1) a Ranking Sensitive PCA and (2) a Ranking Sensitive Vocabulary Boosting. Both schemes embed location cues to learn a compact descriptor, which minimizes the retrieval ranking loss by replacing the original high-dimensional signatures. Our second contribution is a location aware online vocabulary adaption: We store a single vocabulary in the mobile end, which is efficiently adapted for a region specific LDVC coding once a mobile device enters a given region. The learnt LDVC landmark descriptor is extremely compact (typically 10-50 bits with arithmetical coding) and performs superior over state-of-the-art descriptors. We implemented the framework in a real-world mobile landmark search prototype, which is validated in a million-scale landmark database covering typical areas e.g. Beijing, New York City, Lhasa, Singapore, and Florence. [ABSTRACT FROM AUTHOR]