Search

Vision-language navigation is a task that requires an agent to follow instructions to navigate in environments. It becomes increasingly crucial in the field of embodied AI, with potential applications in autonomous navigation, search and rescue, and human-robot interaction. In this paper, we propose to address a more practical yet challenging counterpart setting - vision-language navigation in continuous environments (VLN-CE). To develop a robust VLN-CE agent, we propose a new navigation framework, ETPNav, which focuses on two critical skills: 1) the capability to abstract environments and generate long-range navigation plans, and 2) the ability of obstacle-avoiding control in continuous environments. ETPNav performs online topological mapping of environments by self-organizing predicted waypoints along a traversed path, without prior environmental experience. It privileges the agent to break down the navigation procedure into high-level planning and low-level control. Concurrently, ETPNav utilizes a transformer-based cross-modal planner to generate navigation plans based on topological maps and instructions. The plan is then performed through an obstacle-avoiding controller that leverages a trial-and-error heuristic to prevent navigation from getting stuck in obstacles. Experimental results demonstrate the effectiveness of the proposed method. ETPNav yields more than 10% and 20% improvements over prior state-of-the-art on R2R-CE and RxR-CE datasets, respectively. Our code is available at https://github.com/MarSaKi/ETPNav., Comment: Project page: https://github.com/MarSaKi/ETPNav

Gravitational lensing effect at higher order under weak field approximation is believed to be important to distinguish black holes and other compact objects like wormholes. The deflection angle of a generic wormhole is difficult to solve analytically, thus approximation methods are implemented. In this paper, we investigate the weak-field deflection angle of a specific wormhole, the Ellis-Bronnikov wormhole (EBWH), up to the 1/b^4 order. We use different approximation formalism, study their precision at 1/b^4 order by a comparison to a purely numerical result, and finally rank these formalism by their accuracy. Moreover, we find that certain formalism are sensitive to the choice of coordinate system, when the corresponding deflection angle approaches 0 in the negative-mass branch of universe., Comment: Comments are welcomed

This report presents the methods of the winning entry of the RxR-Habitat Competition in CVPR 2022. The competition addresses the problem of Vision-and-Language Navigation in Continuous Environments (VLN-CE), which requires an agent to follow step-by-step natural language instructions to reach a target. We present a modular plan-and-control approach for the task. Our model consists of three modules: the candidate waypoints predictor (CWP), the history enhanced planner and the tryout controller. In each decision loop, CWP first predicts a set of candidate waypoints based on depth observations from multiple views. It can reduce the complexity of the action space and facilitate planning. Then, a history-enhanced planner is adopted to select one of the candidate waypoints as the subgoal. The planner additionally encodes historical memory to track the navigation progress, which is especially effective for long-horizon navigation. Finally, we propose a non-parametric heuristic controller named tryout to execute low-level actions to reach the planned subgoal. It is based on the trial-and-error mechanism which can help the agent to avoid obstacles and escape from getting stuck. All three modules work hierarchically until the agent stops. We further take several recent advances of Vision-and-Language Navigation (VLN) to improve the performance such as pretraining based on large-scale synthetic in-domain dataset, environment-level data augmentation and snapshot model ensemble. Our model won the RxR-Habitat Competition 2022, with 48% and 90% relative improvements over existing methods on NDTW and SR metrics respectively., Comment: Winner of the 2nd RxR-Habitat Competition @ CVPR2022

The foundation models have recently shown excellent performance on a variety of downstream tasks in computer vision. However, most existing vision foundation models simply focus on image-level pretraining and adpation, which are limited for dynamic and complex video-level understanding tasks. To fill the gap, we present general video foundation models, InternVideo, by taking advantage of both generative and discriminative self-supervised video learning. Specifically, InternVideo efficiently explores masked video modeling and video-language contrastive learning as the pretraining objectives, and selectively coordinates video representations of these two complementary frameworks in a learnable manner to boost various video applications. Without bells and whistles, InternVideo achieves state-of-the-art performance on 39 video datasets from extensive tasks including video action recognition/detection, video-language alignment, and open-world video applications. Especially, our methods can obtain 91.1% and 77.2% top-1 accuracy on the challenging Kinetics-400 and Something-Something V2 benchmarks, respectively. All of these results effectively show the generality of our InternVideo for video understanding. The code will be released at https://github.com/OpenGVLab/InternVideo ., Comment: technical report

Geometric deep learning has been revolutionizing the molecular modeling field. Despite the state-of-the-art neural network models are approaching ab initio accuracy for molecular property prediction, their applications, such as drug discovery and molecular dynamics (MD) simulation, have been hindered by insufficient utilization of geometric information and high computational costs. Here we propose an equivariant geometry-enhanced graph neural network called ViSNet, which elegantly extracts geometric features and efficiently models molecular structures with low computational costs. Our proposed ViSNet outperforms state-of-the-art approaches on multiple MD benchmarks, including MD17, revised MD17 and MD22, and achieves excellent chemical property prediction on QM9 and Molecule3D datasets. Additionally, ViSNet achieved the top winners of PCQM4Mv2 track in the OGB-LCS@NeurIPS2022 competition. Furthermore, through a series of simulations and case studies, ViSNet can efficiently explore the conformational space and provide reasonable interpretability to map geometric representations to molecular structures.

In the technical report, we provide our solution for OGB-LSC 2022 Graph Regression Task. The target of this task is to predict the quantum chemical property, HOMO-LUMO gap for a given molecule on PCQM4Mv2 dataset. In the competition, we designed two kinds of models: Transformer-M-ViSNet which is an geometry-enhanced graph neural network for fully connected molecular graphs and Pretrained-3D-ViSNet which is a pretrained ViSNet by distilling geomeotric information from optimized structures. With an ensemble of 22 models, ViSNet Team achieved the MAE of 0.0723 eV on the test-challenge set, dramatically reducing the error by 39.75% compared with the best method in the last year competition.

Molecular dynamics is a powerful simulation tool to explore material properties. Most of the realistic material systems are too large to be simulated with first-principles molecular dynamics. Classical molecular dynamics has lower computational cost but requires accurate force fields to achieve chemical accuracy. In this work, we develop a symmetry-adapted graph neural networks framework, named molecular dynamics graph neural networks (MDGNN), to construct force fields automatically for molecular dynamics simulations for both molecules and crystals. This architecture consistently preserves the translation, rotation and permutation invariance in the simulations. We propose a new feature engineering method including higher order contributions and show that MDGNN accurately reproduces the results of both classical and first-principles molecular dynamics. We also demonstrate that force fields constructed by the model has good transferability. Therefore, MDGNN provides an efficient and promising option for molecular dynamics simulations of large scale systems with high accuracy.

This paper considers the calibration of Low-Power-Factor wattmeters in high voltage transformer no-load and load loss tests. The value of lowest power factor calibrated is evaluated in consideration of on-site test data and ability of uncertainty evaluation. The harmonic components of both testing current and voltage are measured in the field loss measurement. By analyzing the harmonic components, a nonsinusoidal current with a percentage of 3rd, 5th, 7th harmonic wave is applied for the laboratory calibration of Low-Power-Factor wattmeter in simulation of its on-site tests. There is a voltage unbalance in the field loss measurement. Unbalanced voltage is applied for Low-Power-Factor Wattmeters in comparison with balanced situations. For higher accuracy and resolution, the calibration systems of Low-Power-Factor wattmeters are improved with a nonsinusoidal current and unbalanced voltage standard.

This paper, inspired by mbtiles format which was provided by MapBox, proposes a method on copy of massive map tiles by combing the whole web map tiles into a series of single huge files. Firstly, the map tiles of the same zoom level are packed into a single map file and head file with the extension of .tiles and .tilehead respectively; then create a single sqlite database named sqlint.db3 by the all .tilehead files to provide high efficiently random accessing. Experiments, take the Google Map Tiles as examples, proves good results on copying the tiles into SD cards for mobiles phones. And also the random accessing of tiles from file system directly and from the huge tiles file are also proved that the method has no obvious difference.

Single-crystal diamond (SCD) detectors promise to have applications in neutron spectrometers and fusion neutron monitoring under high flux deuterium plasma. The response to 14 MeV neutrons for the SCD detector is studied in this paper. A high-performance SCD neutron detector is developed by processing cleaning wafer, depositing metal electrodes, annealing and wire-bonding. A fast-neutrons monitoring system containing the detector, preamplifier and digital multichannel analyzer is constructed, and the response to 14 MeV neutrons for the detector is measured on the K-400 neutron generator supported by China Academy of Engineering Physics. In addition, computational simulations of the energy deposition and detection efficiency of 14 MeV neutron through diamond are performed via Geant4 toolkit based on evaluated nuclear data libraries of ENDF-VIII.0, JEFF-3.3, BROND-3.1, JENDL-4.0u and CENDL-3.1. The methods of widening the simulation spectrum and calibration of measuring spectrum are presented in order that simulation results are in reasonable agreement with measured values. The results indicate that the energy deposition of 14 MeV neutrons incident on the 12C can be more accurately calculated with CENDL-3.1 than with other data libraries. The elastic scattering and reaction of 12C(n, 3α) are described more accurately with the CENDL-3.1, and the characteristic peaks of 12C(n, α)9Be matched well the calibrated testing spectrum and the after-widening simulation spectrum, with a difference between the simulated detection efficiency and measuring results being as low as 0.61%. The outcome measures are described as the standardized mean difference, with a detection efficiency of (3.31 × 10–4 ± 0.11 × 10–4) counts/n, an energy resolution of 4.02% ± 0.09%, and a peaking channel of 1797.24 ± 0.80, which suggest that the detector keeps stable well under a high neutron flux of 2 × 1010 n/s for as long as 2 h. The results demonstrate that the SCD detector can be a promising candidate for monitoring 14 MeV D-T neutrons.

To understand the infection status and gene types ofThe fecal samples were collected from 285 HIV/AIDS cases in Nanning, Guilin, Qinzhou, Baise, Hechi cities of Guangxi and 150 HIV negative persons in Nanning City. The modified acid-fast staining and nested-PCR based on 18S rRNA were employed to detect the infection status ofThe infection rate of

In this paper, we present a method using AI techniques to solve a case of pure mathematics applications for finding narrow admissible tuples. The original problem is formulated into a combinatorial optimization problem. In particular, we show how to exploit the local search structure to formulate the problem landscape for dramatic reductions in search space and for non-trivial elimination in search barriers, and then to realize intelligent search strategies for effectively escaping from local minima. Experimental results demonstrate that the proposed method is able to efficiently find best known solutions. This research sheds light on exploiting the local problem structure for an efficient search in combinatorial landscapes as an application of AI to a new problem domain., Comment: 7 pages, 2 figures, conference

To observe the effects of moxibustion and treadmill exercise on transcutaneous oxygen tension and exercise capacity in lower limbs of peripheral arterial disease (PAD).Totally 58 mild-to-moderate PAD patients were assigned to the control group (18 cases), the moxibustion group (20 cases), and the treadmill exercise group (20 cases) by random digit table. Patients in the control group received conventional drug therapy for 12 weeks. Patients in the moxibustion group and the treadmill exercise group additionally received moxibustion [at Zusanli (ST36), Sanyinjiao (SP6), Yongquan (KI1)] and treadmill exercise respectively, once per day, 5 times per week for 12 weeks in total. Ankle-Brachial Index (ABI) , transcutaneous oxygen tension (TcPO₂), 6-min walking test (6MWT), and walking impairment questionnaire (WIQ) were assessed before and after treatment.Compared with the control group and the same group before treatment, there was no statistical difference in ABI in the moxibustion group and the treadmill exercise group (P0.05). But TcPO₂, 6MWT, and WIQ were obviously elevated (P0.01). Besides, 6MWT and WIQ assessment of the treadmill exercise group were better than that of the moxibustion group (P0.01) after intervention.Moxibustion and treadmill exercise could improve the exercise capacity and TcPO₂of lower limbers in PAD patients.

Center-cracked aluminum plates were repaired with single-sided composite patches. The effects of accelerated ageing test on the static and fatigue properties of the specimens were studied. The results show that its mechanical properties are slightly affected by the accelerated one-cycle ageing test. The static bearing capability of the aged specimen is lowered by 2.5%. The fatigue lives of the repaired and aged specimens increased by 8.2 and 8.8 times, respectively, as compared with that of the unrepaired ones.

Objectives To evaluate the effect of effect of Yellow Capsicum extract (YCE) that is rich in capsaicin on the proliferation and differentiation of 3T3-L1 preadipocytes in vitro. Methods 3T3 L1 cells that were exposed to differentiation-inducing medium containing high glucose DMEM (Dulbecco's Modified Eagle’s Medium) and subsequently were treated with capsaicin and YCE for their effect on adipocyte differentiation, changes in their triglyceride content, leptin secretion, expression of lipoprotein lipase, PPARγ, and CCAAT/enhancer-binding protein alpha (C/EBPα). Results Both YCE and capsaicin inhibited proliferation and differentiation 3T3-L1 preadipocytes and suppressed accumulation of intracellular triglyceride in a dose-dependent manner. In addition, a significant decrease in the expression of lipoprotein lipase (LPL), leptin, PPARγ, and C/EBPα was noted in 3T3-L1 preadipocytes when induced to differentiate by YCE and Capsaicin. Conclusions The potent inhibitory action of YCE and Capsaicin on the differentiation of 3T3-L1 preadipocyte observed suggests that they (YCE and Capsaicin) have the potential to inhibit obesity that needs to be explored in future studies.

A modification of the transition state theory for evaporation and condensation is presented by analysing the kinetic characteristics of liquid-vapour interphase transport. In the modified transition state theory, the moving orientation of molecules is introduced into the calculation of the free volume of the activated complex. The condensation coefficients of argon at different temperatures are calculated with the modified transition state theory. The results agree well with those from molecular dynamics simulations.

It has been proved in a long period of practice that the planting of rice is effective on the improvement of saline soils. This can be applied to the extremely arid areas in southern Xinjiang(l). Flooding in the rice growing season makes it possible for the irrigation water to dissolve salt and thence remove it in drainage. In the meantime, the salt content of underground water in paddy fields is gradually on the decrease. Even in places where drainage conditions are unfavorable, such practices as irrigation, drainage, fertilization, and rotation of upland crop and paddy rice are still beneficial to the control of salt content in the soil and an increase of crop yield is expected.