Recently, a state-of-the-art algorithm, called deep deterministic policy gradient (DDPG), has achieved good performance in many continuous control tasks the MuJoCo simulator. To further improve efficiency of experience replay mechanism DDPG and thus speeding up training process, this paper, prioritized method is proposed for where sampling adopted instead uniform sampling. The with tested an inverted pendulum task via OpenAI Gym. experimental results show that can reduce time stability less...

Sounds provide rich semantics, complementary to visual data, for many tasks. However, in practice, sounds from multiple sources are often mixed together. In this paper we propose a novel framework, referred as MinusPlus Network (MP-Net), the task of sound separation. MP-Net separates recursively order average energy, removing separated mixture at end each prediction, until becomes empty or contains only noise. way, could be applied mixtures with arbitrary numbers and types sounds. Moreover,...

Multi-modality perception is essential to develop interactive intelligence. In this work, we consider a new task of visual information-infused audio inpainting, i.e. synthesizing missing segments that correspond their accompanying videos. We identify two key aspects for successful inpainter: (1) It desirable operate on spectrograms instead raw audios. Recent advances in deep semantic image inpainting could be leveraged go beyond the limitations traditional inpainting. (2) To synthesize...

Stereophonic audio, especially binaural plays an essential role in immersive viewing environments. Recent research has explored generating visually guided stereophonic audios supervised by multi-channel audio collections. However, due to the requirement of professional recording devices, existing datasets are limited scale and variety, which impedes generalization methods real-world scenarios. In this work, we propose PseudoBinaural, effective pipeline that is free recordings. The key...

3D point cloud is an important representation for capturing real world objects. However, real-scanned clouds are often incomplete, and it to recover complete downstream applications. Most existing completion methods use Chamfer Distance (CD) loss training. The CD estimates correspondences between two by searching nearest neighbors, which does not capture the overall density distribution on generated shape, therefore likely leads non-uniform generation. To tackle this problem, we propose a...

Neural surface reconstruction aims to reconstruct accurate 3D surfaces based on multi-view images. Previous methods neural volume rendering mostly train a fully implicit model with MLPs, which typically require hours of training for single scene. Recent efforts explore the explicit volumetric representation accelerate optimization via memorizing significant information learnable voxel grids. However, existing voxel-based often struggle in reconstructing fine-grained geometry, even when...

The first case of a decrease in density with the introduction trinitromethyl group.

Full nitration is one of the most effective strategies used in synthesizing high-density energetic materials, but this strategy has reached its limit because resultant compounds cannot be further functionalized. To overcome limitation, we present synergistic action full and strong intermolecular H-bonding designing 1-trinitromethyl-3,5-dinitro-4-nitroaminopyrazole (DNTP) with a density that exceeds those reported monocyclic CHON compounds. The detonation velocity specific impulse DNTP exceed...

In this study, we propose a "modular assembly" strategy to synthesize series of new trinitromethyl high-energy molecules. For typical molecule 1, approach reduces the number synthetic steps from 7 2 and improves yield 1.0% 48.3%, representing more than 48-fold increase compared traditional "skeleton first, group later" method. All synthesized molecules exhibit exceptional energetic performance, with 1 demonstrating density 1.913 g cm–3, detonation velocity 9151 m s–1, thermal decomposition...

The advancement of generative radiance fields has pushed the boundary 3D-aware image synthesis. Motivated by observation that a 3D object should look realistic from multiple viewpoints, these methods introduce multi-view constraint as regularization to learn valid 2D images. Despite progress, they often fall short capturing accurate shapes due shape-color ambiguity, limiting their applicability in downstream tasks. In this work, we address ambiguity proposing novel shading-guided implicit...

To improve the forming quality and limit of numerical control (NC) bending high-pressure titanium alloy tubes, in this study, using three-dimensional (3D) finite element method, deformation behavior medium-strength TA18 tubes during NC with different radii is investigated. The results show that cross-sectional wall thickness variation a small radius (less than 2 times tube outside diameter) are clearly from normal (between 4 diameter). For radius, or without mandrel, distribution flattening...

The advent of generative radiance fields has significantly promoted the development 3D-aware image synthesis. cumulative rendering process in makes training these models much easier since gradients are distributed over entire volume, but leads to diffused object surfaces. In meantime, compared occupancy representations could inherently ensure deterministic However, if we directly apply models, during they will only receive sparse located on surfaces and eventually suffer from convergence...

Inner wrinkling phenomenon is more likely to develop during hydrodynamic deep drawing (HDD) of complicated component-forms due the higher demand for controlling deformation sequences. Aiming at problems in control inner an irregular surface part featured with both concavity and convex, this research proposes optimal design method drawbead parameters change material flow. According theoretical analysis mechanism wrinkling, optimizing cavity pressure only unreasonable form a wrinkle-free...

The noise control of flank array sonar is a primary approach to enhance the detection range. During submarine navigation, hydrodynamic main source in platform region sonar, which includes flow and flow-induced noise. Therefore, an in-depth investigation necessary. In this paper, we firstly take teardrop as computational model validate method. Afterwards, numerically simulate characteristics for cylindrical shell model, investigate differences at different points along X, Y, Z axes. Finally,...

Stereophonic audio is an indispensable ingredient to enhance human auditory experience. Recent research has explored the usage of visual information as guidance generate binaural or ambisonic from mono ones with stereo supervision. However, this fully supervised paradigm suffers inherent drawback: recording stereophonic usually requires delicate devices that are expensive for wide accessibility. To overcome challenge, we propose leverage vastly available data facilitate generation audio. Our...

Based on powerful text-to-image diffusion models, text-to-3D generation has made significant progress in generating compelling geometry and appearance. However, existing methods still struggle to recover high-fidelity object materials, either only considering Lambertian reflectance, or failing disentangle BRDF materials from the environment lights. In this work, we propose Material-Aware Text-to-3D via LAtent auto-EncodeR (\textbf{MATLABER}) that leverages a novel latent auto-encoder for...