A5076586872- Perovskite Materials and Applications
- Advanced Semiconductor Detectors and Materials
- Chalcogenide Semiconductor Thin Films
- Solid-state spectroscopy and crystallography
- Radiation Detection and Scintillator Technologies
- Advanced Neural Network Applications
- Luminescence Properties of Advanced Materials
- Domain Adaptation and Few-Shot Learning
- Optical properties and cooling technologies in crystalline materials
- Crystallization and Solubility Studies
- X-ray Diffraction in Crystallography
- Semiconductor Quantum Structures and Devices
- Crystal Structures and Properties
- Video Surveillance and Tracking Methods
- Quantum Dots Synthesis And Properties
- 2D Materials and Applications
- Human Pose and Action Recognition
- Advanced X-ray and CT Imaging
- Machine Learning in Materials Science
- Organic Light-Emitting Diodes Research
- Luminescence and Fluorescent Materials
- Adversarial Robustness in Machine Learning
- Organic Electronics and Photovoltaics
- Anomaly Detection Techniques and Applications
- Advanced Photocatalysis Techniques
Soochow University
2021-2025
East China Normal University
2023-2025
Carnegie Mellon University
2018-2024
Northwestern University
2017-2024
Michigan State University
2023
Sichuan Agricultural University
2023
University of Cagliari
2023
Columbia University
2023
Wuhan Institute of Technology
2018-2019
Xi'an Jiaotong University
2017-2018
In this paper, we introduce a new channel pruning method to accelerate very deep convolutional neural networks. Given trained CNN model, propose an iterative two-step algorithm effectively prune each layer, by LASSO regression based selection and least square reconstruction. We further generalize multi-layer multi-branch cases. Our reduces the accumulated error enhance compatibility with various architectures. pruned VGG-16 achieves state-of-the-art results 5× speed-up along only 0.3%...
person 0.83 0.90 0.69 0.68 0.57 skis 0.59 0.53 a: RetinaNet (anchor-based, ResNeXt-101) b: Ours (anchor-based + FSAF, ResNet-50) Figure 1: Qualitative results of the anchor-based [22] using powerful ResNeXt-101 (left) and our detector with additional FSAF module just ResNet-50 (right) under same training testing scale.Our helps detecting hard objects like tiny flat a less backbone network.See 7 for more examples.
Large-scale object detection datasets (e.g., MS-COCO) try to define the ground truth bounding boxes as clear possible. However, we observe that ambiguities are still introduced when labeling boxes. In this paper, propose a novel box regression loss for learning transformation and localization variance together. Our greatly improves accuracies of various architectures with nearly no additional computation. The learned allows us merge neighboring during non-maximum suppression (NMS), which...
Gamma-ray detection and spectroscopy is the quantitative determination of their energy spectra, critical value critically important in diverse technological scientific fields. Here we report an improved melt growth method for cesium lead bromide a special detector design with asymmetrical metal electrode configuration that leads to high performance at room temperature. As-grown centimeter-sized crystals possess extremely low impurity levels (below 10 p.p.m. total 69 elements) detectors...
Abstract The organic-inorganic hybrid lead halide perovskites have emerged as a series of star materials for solar cells, lasers and detectors. However, the issues raised by toxic element marginal stability due to volatile organic components severely limited their potential applications. In this work, we develop nucleation-controlled solution method grow large size high-quality Cs 3 Bi 2 I 9 perovskite single crystals (PSCs). Using technique, harvest some centimeter-sized achieved high...
Model compression is a critical technique to efficiently deploy neural network models on mobile devices which have limited computation resources and tight power budgets. Conventional model techniques rely hand-crafted heuristics rule-based policies that require domain experts explore the large design space trading off among size, speed, accuracy, usually sub-optimal time-consuming. In this paper, we propose AutoML for Compression (AMC) leverage reinforcement learning provide policy. This...
Abstract Low ionic migration is required for a semiconductor material to realize stable high‐performance X‐ray detection. In this work, successful controlled incorporation of not only methylammonium (MA + ) and cesium (Cs cations, but also bromine (Br – anions into the FAPbI 3 lattice grow inch‐sized perovskite single crystal (FAMACs SC) reported. The smaller cations anions, comparing original FA I help release stress so that FAMACs SC shows lower ion migration, enhanced hardness, trap...
The unique hybrid nature of 2D Ruddlesden-Popper (R-P) perovskites has bestowed upon them not only tunability their electronic properties but also high-performance devices with improved environmental stability as compared to 3D analogs. However, there is limited information about inherent heat, light, and air how different parameters such the inorganic layer number length organic spacer molecule affect stability. To gain deeper understanding on matter we have expanded family R-P perovskites,...
In this paper, we introduce a new channel pruning method to accelerate very deep convolutional neural networks.Given trained CNN model, propose an iterative two-step algorithm effectively prune each layer, by LASSO regression based selection and least square reconstruction. We further generalize multi-layer multi-branch cases. Our reduces the accumulated error enhance compatibility with various architectures. pruned VGG-16 achieves state-of-the-art results 5x speed-up along only 0.3%...
The two-dimensional Ruddlesden-Popper (RP) phases are an important class of halide perovskites with versatile optoelectronic properties. So far, only organic-inorganic hybrid RP involving long organic spacers were reported in this class. Here, we report all-inorganic phase lead perovskite, Cs2PbI2Cl2 (1, I4/ mmm space group; a = 5.6385(8) Å, c 18.879(4) Å), synthesized by solid-state method. compound exhibits band gap Eg ∼ 3.04 eV and photoconductivity. We find anomalous evolution Cs2Pb1-...
A common approach to localize 3D human joints in a synchronized and calibrated multi-view setup consists of two-steps: (1) apply 2D detector separately on each view 2D, (2) perform robust triangulation detections from acquire the joint locations. However, step 1, is limited solving challenging cases which could potentially be better resolved 3D, such as occlusions oblique viewing angles, purely without leveraging any information. Therefore, we propose differentiable "epipolar transformer",...
Pressure processing is efficient to regulate the structural and physical properties of two-dimensional (2D) halide perovskites which have been emerging for advanced photovoltaic light-emitting applications. Increasing numbers studies reported pressure-induced and/or enhanced emission in 2D perovskites. However, no research has focused on their photoresponse under pressure tuning. It also unclear how change affects excitonic features, govern optoelectronic Herein, we report significantly...
Halide perovskites exhibit remarkably high-performance as semiconductors compared to conventional materials because of an unusually favorable combination optoelectronic properties. We demonstrate here that solution-grown single-crystals organic–inorganic hybrid perovskite CH3NH3PbI3 (MAPbI3), implemented in a Schottky-type device design, can produce outstanding hard radiation detectors with high spectral response and low dark current for the first time. MAPbI3 detector achieves excellent...
Large organic A cations cannot stabilize the 3D perovskite AMX3 structure because they be accommodated in cubo-octhedral cage (do not follow Goldschmidt tolerance factor rule), and generally template low-dimensional structures. Here we report that large dication aminomethylpyridinium (AMPY) can novel structures which resemble conventional perovskites. They have formula (xAMPY)M2I6 (x = 3 or 4, M Sn2+ Pb2+) is double of formula. However, steric requirement rule, it impossible for to form...
We have investigated the defect perovskites A3M2I9 (A = Cs, Rb; M Bi, Sb) as materials for radiation detection. The phase purity of Bridgman-grown single crystals was confirmed via high-resolution synchrotron X-ray diffraction, while density functional theory calculations (DFT) show surprisingly dispersive bands in out-of-plane direction these layered materials, with low effective masses both holes and electrons. Accordingly, each four showed response to 241Am α-particle irradiation hole...
The detection of γ-rays at room temperature with high-energy resolution using semiconductors is one the most challenging applications. presence even smallest amount defects sufficient to kill signal generated from which makes availability detectors a rarity. Lead halide perovskite exhibit unusually high defect tolerance leading outstanding and unique optoelectronic properties are poised strongly impact applications in photoelectric conversion/detection. Here we demonstrate for first time...
Hybrid organic-inorganic halide perovskites have shown remarkable optoelectronic properties, exhibiting an impressive tolerance to defects believed originate from correlated motion of charge carriers and the polar lattice forming large polarons. Few experimental techniques are capable directly probing these correlations, requiring simultaneous sub-millielectron volt energy femtosecond temporal resolution after absorption a photon. Here, we use time-resolved multi-THz spectroscopy, sensitive...
MobileNet and Binary Neural Networks are two among the most widely used techniques to construct deep learning models for performing a variety of tasks on mobile embedded platforms. In this paper, we present simple yet efficient scheme exploit binarization at activation function model weights. However, training binary network from scratch with separable depth-wise point-wise convolutions in case is not trivial prone divergence. To tackle issue, propose novel neural architecture, namely MoBi-...
Abstract Spectroscopic‐grade single crystal detectors can register the energies of individual X‐ray interactions enabling photon‐counting systems with superior resolution over traditional photoconductive detection systems. Current technical challenges have limited preparation perovskite semiconductors for energy‐discrimination detection. Here, this work reports deployment a spectroscopic‐grade CsPbBr 3 Schottky detector under reverse bias continuum hard in both photocurrent and spectroscopic...