The irregular domain and lack of ordering make it challenging to design deep neural networks for point cloud processing. This paper presents a novel framework named
Point Cloud Transformer
(PCT) for ...point cloud learning. PCT is based on Transformer, which achieves huge success in natural language processing and displays great potential in image processing. It is inherently permutation invariant for processing a sequence of points, making it well-suited for point cloud learning. To better capture local context within the point cloud, we enhance input embedding with the support of farthest point sampling and nearest neighbor search. Extensive experiments demonstrate that the PCT achieves the state-of-the-art performance on shape classification, part segmentation, semantic segmentation, and normal estimation tasks.
Attention mechanisms, especially self-attention, have played an increasingly important role in deep feature representation for visual tasks. Self-attention updates the feature at each position by ...computing a weighted sum of features using pair-wise affinities across all positions to capture the long-range dependency within a single sample. However, self-attention has quadratic complexity and ignores potential correlation between different samples. This article proposes a novel attention mechanism which we call external attention , based on two external, small, learnable, shared memories, which can be implemented easily by simply using two cascaded linear layers and two normalization layers; it conveniently replaces self-attention in existing popular architectures. External attention has linear complexity and implicitly considers the correlations between all data samples. We further incorporate the multi-head mechanism into external attention to provide an all-MLP architecture, external attention MLP (EAMLP), for image classification. Extensive experiments on image classification, object detection, semantic segmentation, instance segmentation, image generation, and point cloud analysis reveal that our method provides results comparable or superior to the self-attention mechanism and some of its variants, with much lower computational and memory costs.
Humans can naturally and effectively find salient regions in complex scenes. Motivated by this observation, attention mechanisms were introduced into computer vision with the aim of imitating this ...aspect of the human visual system. Such an attention mechanism can be regarded as a dynamic weight adjustment process based on features of the input image. Attention mechanisms have achieved great success in many visual tasks, including image classification, object detection, semantic segmentation, video understanding, image generation, 3D vision, multimodal tasks, and self-supervised learning. In this survey, we provide a comprehensive review of various attention mechanisms in computer vision and categorize them according to approach, such as channel attention, spatial attention, temporal attention, and branch attention; a related repository
https://github.com/MenghaoGuo/Awesome-Vision-Attentions
is dedicated to collecting related work. We also suggest future directions for attention mechanism research.
The long noncoding RNA NEAT1 (nuclear enriched abundant transcript 1) nucleates the formation of paraspeckles, which constitute a type of nuclear body with multiple roles in gene expression. Here we ...identify NEAT1 regulators using an endogenous NEAT1 promoter-driven enhanced green fluorescent protein reporter in human cells coupled with genome-wide RNAi screens. The screens unexpectedly yield gene candidates involved in mitochondrial functions as essential regulators of NEAT1 expression and paraspeckle formation. Depletion of mitochondrial proteins and treatment of mitochondrial stressors both lead to aberrant NEAT1 expression via ATF2 as well as altered morphology and numbers of paraspeckles. These changes result in enhanced retention of mRNAs of nuclear-encoded mitochondrial proteins (mito-mRNAs) in paraspeckles. Correspondingly, NEAT1 depletion has profound effects on mitochondrial dynamics and function by altering the sequestration of mito-mRNAs in paraspeckles. Overall, our data provide a rich resource for understanding NEAT1 and paraspeckle regulation, and reveal a cross-regulation between paraspeckles and mitochondria.
Global Contrast Based Salient Region Detection Cheng, Ming-Ming; Mitra, Niloy J.; Huang, Xiaolei ...
IEEE transactions on pattern analysis and machine intelligence,
2015-March-1, 2015-Mar, 2015-3-1, 20150301, Letnik:
37, Številka:
3
Journal Article
Recenzirano
Odprti dostop
Automatic estimation of salient object regions across images, without any prior assumption or knowledge of the contents of the corresponding scenes, enhances many computer vision and computer ...graphics applications. We introduce a regional contrast based salient object detection algorithm, which simultaneously evaluates global contrast differences and spatial weighted coherence scores. The proposed algorithm is simple, efficient, naturally multi-scale, and produces full-resolution, high-quality saliency maps. These saliency maps are further used to initialize a novel iterative version of GrabCut, namely SaliencyCut, for high quality unsupervised salient object segmentation. We extensively evaluated our algorithm using traditional salient object detection datasets, as well as a more challenging Internet image dataset. Our experimental results demonstrate that our algorithm consistently outperforms 15 existing salient object detection and segmentation methods, yielding higher precision and better recall rates. We also show that our algorithm can be used to efficiently extract salient object masks from Internet images, enabling effective sketch-based image retrieval (SBIR) via simple shape comparisons. Despite such noisy internet images, where the saliency regions are ambiguous, our saliency guided image retrieval achieves a superior retrieval rate compared with state-of-the-art SBIR methods, and additionally provides important target object region information.
Photorealistic Audio-driven Video Portraits Wen, Xin; Wang, Miao; Richardt, Christian ...
IEEE transactions on visualization and computer graphics,
12/2020, Letnik:
26, Številka:
12
Journal Article
Recenzirano
Odprti dostop
Video portraits are common in a variety of applications, such as videoconferencing, news broadcasting, and virtual education and training. We present a novel method to synthesize photorealistic video ...portraits for an input portrait video, automatically driven by a person's voice. The main challenge in this task is the hallucination of plausible, photorealistic facial expressions from input speech audio. To address this challenge, we employ a parametric 3D face model represented by geometry, facial expression, illumination, etc., and learn a mapping from audio features to model parameters. The input source audio is first represented as a high-dimensional feature, which is used to predict facial expression parameters of the 3D face model. We then replace the expression parameters computed from the original target video with the predicted one, and rerender the reenacted face. Finally, we generate a photorealistic video portrait from the reenacted synthetic face sequence via a neural face renderer. One appealing feature of our approach is the generalization capability for various input speech audio, including synthetic speech audio from text-to-speech software. Extensive experimental results show that our approach outperforms previous general-purpose audio-driven video portrait methods. This includes a user study demonstrating that our results are rated as more realistic than previous methods.
We investigate how blockchain technology influences optimal pricing and quality decisions under two different supply chain structures. In both cases, the supplier is the leader and plays a ...Stackelberg game with downstream platform(s). Specifically, the supplier determines the wholesale prices and quality levels of the products, and the platform(s) decides the selling prices. Based on our analysis, we find that when consumer acceptance of blockchain is low, among the two products in the single platform, the products with blockchain technology have not only a high quality but also a low price. However, in the case of the two competitive platforms, although blockchain will also enable the platform that uses this technology to provide products with a high quality and low price, its advantages are not obvious. Nevertheless, in comparing the two cases, when consumer acceptance is high, the two competing platforms tend to provide products with a high quality and low price, regardless of whether or not the products use blockchain technology. Our conclusions reveal that although blockchain will benefit a supplier and consumers, in some cases, the supplier will have an incentive to decrease the quality level of the products with blockchain technology, which will eventually negatively affect consumers.
Heteronuclear BeFe(CO)4− anion complex is generated in the gas phase, which is detected by mass‐selected infrared photodissociation spectroscopy in the carbonyl stretching frequency region. The ...complex is characterized to have a Be−Fe bonded Be−Fe(CO)4− structure with C3v symmetry and all of the four carbonyl ligands bonded on the iron center. Quantum chemical studies indicate that the complex has a quite short Be−Fe bond. Besides one electron‐sharing σ bond, there are two additional, albeit weak, Be ← Fe(CO)4− dative π bonding interactions. The findings imply that metal–metal bonding between s‐block and transition metals is viable under suitable coordination environment.
The heteronuclear BeFe(CO)4− anion complex is generated in the gas phase, which is characterized to have a quite short Be−Fe bonded C3v structure. The findings imply that metal–metal bonding between s‐block and transition metals is viable under suitable coordination environment.
Spatially and temporally adaptive algorithms can substantially improve the computational efficiency of many numerical schemes in computational mechanics and physics‐based animation. Recently, a ...crucial need for temporal adaptivity in the Material Point Method (MPM) is emerging due to the potentially substantial variation of material stiffness and velocities in multi‐material scenes. In this work, we propose a novel temporally adaptive symplectic Euler scheme for MPM with regional time stepping (RTS), where different time steps are used in different regions. We design a time stepping scheduler operating at the granularity of small blocks to maintain a natural consistency with the hybrid particle/grid nature of MPM. Our method utilizes the Sparse Paged Grid (SPGrid) data structure and simultaneously offers high efficiency and notable ease of implementation with a practical multi‐threaded particle‐grid transfer strategy. We demonstrate the efficacy of our asynchronous MPM method on various examples including elastic objects, granular media, and fluids.
Seeded vacuum decay with Gauss-Bonnet Gregory, Ruth; Hu, Shi-Qian
The journal of high energy physics,
11/2023, Letnik:
2023, Številka:
11
Journal Article
Recenzirano
Odprti dostop
A
bstract
We investigate false vacuum decay catalysed by black holes under the influence of the higher order Gauss-Bonnet term. We study both bubble nucleation and Hawking-Moss types of phase ...transition in arbitrary dimension. The equations of motion of “bounce” solutions in which bubbles nucleate around arbitrary dimensional black holes are found in the thin wall approximation, and the instanton action is computed. The headline result that the tunnelling action for static instantons is the difference in entropy of the seed and remnant black holes is shown to hold for arbitrary dimension. We also study the Hawking-Moss transition and find a picture similar to the Einstein case, with one curious five-dimensional exception (due to a mass gap). In four dimensions, we find as expected that the Gauss-Bonnet term only impacts topology changing transitions, i.e. when vacuum decay removes the seed black hole altogether, or in a (Hawking-Moss) transition where a black hole is created. In the former case, topology changing transitions are suppressed (for positive GB coupling
α
), whereas the latter case results in an enhanced transition.
PDF
