Search

Learning-based multi-view stereo (MVS) has by far centered around 3D convolution on cost volumes. Due to the high computation and memory consumption of 3D CNN, the resolution of output depth is often considerably limited. Different from most existing works dedicated to adaptive refinement of cost volumes, we opt to directly optimize the depth value along each camera ray, mimicking the range (depth) finding of a laser scanner. This reduces the MVS problem to ray-based depth optimization which is much more light-weight than full cost volume optimization. In particular, we propose RayMVSNet which learns sequential prediction of a 1D implicit field along each camera ray with the zero-crossing point indicating scene depth. This sequential modeling, conducted based on transformer features, essentially learns the epipolar line search in traditional multi-view stereo. We devise a multi-task learning for better optimization convergence and depth accuracy. We found the monotonicity property of the SDFs along each ray greatly benefits the depth estimation. Our method ranks top on both the DTU and the Tanks & Temples datasets over all previous learning-based methods, achieving an overall reconstruction score of 0.33 mm on DTU and an F-score of 59.48% on Tanks & Temples. It is able to produce high-quality depth estimation and point cloud reconstruction in challenging scenarios such as objects/scenes with non-textured surface, severe occlusion, and highly varying depth range. Further, we propose RayMVSNet++ to enhance contextual feature aggregation for each ray through designing an attentional gating unit to select semantically relevant neighboring rays within the local frustum around that ray. This improves the performance on datasets with more challenging examples (e.g., low-quality images caused by poor lighting conditions or motion blur). RayMVSNet++ achieves state-of-the-art performance on the ScanNet dataset. In particular, it attains an AbsRel of 0.058m and produces accurate results on the two subsets of textureless regions and large depth variation.

3D symmetry detection is a fundamental problem in computer vision and graphics. Most prior works detect symmetry when the object model is fully known, few studies symmetry detection on objects with partial observation, such as single RGB-D images. Recent work addresses the problem of detecting symmetries from incomplete data with a deep neural network by leveraging the dense and accurate symmetry annotations. However, due to the tedious labeling process, full symmetry annotations are not always practically available. In this work, we present a 3D symmetry detection approach to detect symmetry from single-view RGB-D images without using symmetry supervision. The key idea is to train the network in a weakly-supervised learning manner to complete the shape based on the predicted symmetry such that the completed shape be similar to existing plausible shapes. To achieve this, we first propose a discriminative variational autoencoder to learn the shape prior in order to determine whether a 3D shape is plausible or not. Based on the learned shape prior, a symmetry detection network is present to predict symmetries that produce shapes with high shape plausibility when completed based on those symmetries. Moreover, to facilitate end-to-end network training and multiple symmetry detection, we introduce a new symmetry parametrization for the learning-based symmetry estimation of both reflectional and rotational symmetry. The proposed approach, coupled symmetry detection with shape completion, essentially learns the symmetry-aware shape prior, facilitating more accurate and robust symmetry detection. Experiments demonstrate that the proposed method is capable of detecting reflectional and rotational symmetries accurately, and shows good generality in challenging scenarios, such as objects with heavy occlusion and scanning noise. Moreover, it achieves state-of-the-art performance, improving the F1-score over the existing supervised learning method by 2%-11% on the ShapeNet and ScanNet datasets.

The synthesis of size-controlled Sm2Fe17magnetic particles is a prerequisite for the fabrication of high-performance Sm2Fe17N3permanent magnetic materials. Here, Sm2Fe17was synthesized using a cost-effective reduction-diffusion method. The calcium chloride molten salt was introduced to control the particle size and achieve a single phase of Sm2Fe17. The effects of reduction-diffusion reaction temperature and the amount of added calcium chloride on the phase constitution and microstructure of the final product of reduction-diffusion were systematically investigated. Adding an appropriate amount of calcium chloride can effectively inhibit the overgrowth and sintering of the reduced particles. By employing the strategy of adding 20 wt% of calcium chlorides into the green compacts, we were able to successfully synthesize uniform Sm2Fe17particles that are well-dispersed, with an average size of 2.2 μm. Furthermore, by combining the optimal reduction-diffusion conditions and the nitriding process, the hard magnetic Sm2Fe17N3material was successfully obtained. This study could be useful for the development of high-performance Sm2Fe17N3magnetic materials utilizing reduction-diffusion technology.

In this study, 3-dimensional nanoflower structured TiO2photoanodes were successfully grown onto tin oxide doped with fluorine substrates. X-ray diffraction and scanning electron microscopy were performed. The 3-dimensional TiO2nanoflowers presented improved photoelectrochemical performance compared with a TiO2layer. Meanwhile, TiO2nanoflowers with 2.5 h preparation time substantially increased the production of H2compared with the TiO2layer. The improved performance could be ascribed to the 3-dimensional nanoflower structure with many carriers and fast reaction kinetics, which enhanced light absorption and enlarged the contact area between the photoanode and electrolyte. Moreover, this nanoflower structure with trunks connected to the FTO substrate also improved the transfer of photogenerated carriers.

Deformability and compact energy density are urgently needed for power supplies to cater to today's increasingly multifunctional consumer electronics. Here, a flexible and self-supporting composite electrode, in which manganese oxide directly grows on surfaces of the reduced graphene oxide and the interconnected porous framework of the hydrogel is retained, was fabricated by electrochemical deposition of manganese dioxide in graphene hydrogel. Thanks to the contribution of pseudocapacitance and the unchanged volume of the hydrogel, the introduction of manganese oxide helps to lift the volumetric capacitance of the electrode by more than 400%, from 103.6 to 417.8 F cm[?]3. The solid-state asymmetric supercapacitor, composed of an activated carbon electrode and an as-prepared composite electrode, has a capacitance of 67.7 F cm[?]3. The series connected supercapacitor not only successfully illuminates a white light-emitting diode bulb, but also works steadily under deformation conditions.

Compositing TiO2with metal sulfide to construct heterojunction is an effective approach to improve the carriers’ separation efficiency and photocatalytic activity. However, TiO2is an N-type semiconductor and most metal sulfide is also N-type semiconductor. It will form n–n heterojunction between TiO2and metal sulfide. The photoinduced electrons in conduction band of metal sulfide hardly flow into the conduction band of TiO2, which will weaken the improvement of carriers’ separation efficiency. In this work, anatase TiO2nanosheets with coexposed {101} and {001} facets were composited with porous ZnS to construct a novel n–p–n dual heterojunction. This TiO2/ZnS composite displays 108% improvement of photocatalytic activity compared to that of pristine TiO2nanosheets. As comparison, P25 with mainly exposed {101} facets/porous ZnS with an n–n single heterojunction only show 2% enhancement of photocatalytic activity than that of P25. The n–p–n dual heterojunction displays an obvious advantage than common TiO2/ZnS n–n heterojunction. In the n–p–n dual heterojunction, first, photoinduced electrons at CB of {001} facets will flow into the CB of {101} facets, while photoinduced holes at VB of {101} facets will flow into the VB of {001} facets; second, photoinduced electrons at CB of ZnS will flow into the CB of {001} facets, while photoinduced holes at VB of {001} facets will flow into the VB of ZnS. In this way, it realizes carriers’ separation in the n–p–n dual heterojunction. This work improves a new strategy to employ crystal facets of photocatalysts to construct n–p–n dual heterojunction with metal sulfide for enhancing the photocatalytic activity.

Deformability and compact energy density are urgently needed for power supplies to cater to today's increasingly multifunctional consumer electronics. Here, a flexible and self-supporting composite electrode, in which manganese oxide directly grows on surfaces of the reduced graphene oxide and the interconnected porous framework of the hydrogel is retained, was fabricated by electrochemical deposition of manganese dioxide in graphene hydrogel. Thanks to the contribution of pseudocapacitance and the unchanged volume of the hydrogel, the introduction of manganese oxide helps to lift the volumetric capacitance of the electrode by more than 400%, from 103.6 to 417.8 F cm−3. The solid-state asymmetric supercapacitor, composed of an activated carbon electrode and an as-prepared composite electrode, has a capacitance of 67.7 F cm−3. The series connected supercapacitor not only successfully illuminates a white light-emitting diode bulb, but also works steadily under deformation conditions.