Search

Your search keyword '"Li, Congcong"' showing total 93 results
Start Over Author "Li, Congcong" Publisher ieee
93 results on '"Li, Congcong"'

15. Multi-modal 3D Human Pose Estimation with 2D Weak Supervision in Autonomous Driving

Author

Zheng, Jingxiao, Shi, Xinwei, Gorban, Alexander, Mao, Junhua, Song, Yang, Qi, Charles R., Liu, Ting, Chari, Visesh, Cornman, Andre, Zhou, Yin, Li, Congcong, and Anguelov, Dragomir

Subjects
FOS: Computer and information sciences, Computer Vision and Pattern Recognition (cs.CV), ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION, Computer Science - Computer Vision and Pattern Recognition
Abstract

3D human pose estimation (HPE) in autonomous vehicles (AV) differs from other use cases in many factors, including the 3D resolution and range of data, absence of dense depth maps, failure modes for LiDAR, relative location between the camera and LiDAR, and a high bar for estimation accuracy. Data collected for other use cases (such as virtual reality, gaming, and animation) may therefore not be usable for AV applications. This necessitates the collection and annotation of a large amount of 3D data for HPE in AV, which is time-consuming and expensive. In this paper, we propose one of the first approaches to alleviate this problem in the AV setting. Specifically, we propose a multi-modal approach which uses 2D labels on RGB images as weak supervision to perform 3D HPE. The proposed multi-modal architecture incorporates LiDAR and camera inputs with an auxiliary segmentation branch. On the Waymo Open Dataset, our approach achieves a 22% relative improvement over camera-only 2D HPE baseline, and 6% improvement over LiDAR-only model. Finally, careful ablation studies and parts based analysis illustrate the advantages of each of our contributions.

Published
2022

16. End-to-End Compressed Video Representation Learning for Generic Event Boundary Detection

Author

Li, Congcong, Wang, Xinyao, Wen, Longyin, Hong, Dexiang, Luo, Tiejian, and Zhang, Libo

Subjects
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION, Computer Science - Computer Vision and Pattern Recognition
Abstract

Generic event boundary detection aims to localize the generic, taxonomy-free event boundaries that segment videos into chunks. Existing methods typically require video frames to be decoded before feeding into the network, which demands considerable computational power and storage space. To that end, we propose a new end-to-end compressed video representation learning for event boundary detection that leverages the rich information in the compressed domain, i.e., RGB, motion vectors, residuals, and the internal group of pictures (GOP) structure, without fully decoding the video. Specifically, we first use the ConvNets to extract features of the I-frames in the GOPs. After that, a light-weight spatial-channel compressed encoder is designed to compute the feature representations of the P-frames based on the motion vectors, residuals and representations of their dependent I-frames. A temporal contrastive module is proposed to determine the event boundaries of video sequences. To remedy the ambiguities of annotations and speed up the training process, we use the Gaussian kernel to preprocess the ground-truth event boundaries. Extensive experiments conducted on the Kinetics-GEBD dataset demonstrate that the proposed method achieves comparable results to the state-of-the-art methods with $4.5\times$ faster running speed.

Published
2022

21. Depth Estimation Matters Most: Improving Per-Object Depth Estimation for Monocular 3D Detection and Tracking

Author

Jing, Longlong, primary, Yu, Ruichi, additional, Kretzschmar, Henrik, additional, Li, Kang, additional, Qi, Charles R., additional, Zhao, Hang, additional, Ayvaci, Alper, additional, Chen, Xu, additional, Cower, Dillon, additional, Li, Yingwei, additional, You, Yurong, additional, Deng, Han, additional, Li, Congcong, additional, and Anguelov, Dragomir, additional

Published
2022
Full Text
View/download PDF

30. A Time-Domain Computing-in-Memory based Processor using Predictable Decomposed Convolution for Arbitrary Quantized DNNs

Author

Yang, Jianxun, primary, Kong, Yuyao, additional, Zhang, Zhao, additional, Liu, Zhuangzhi, additional, Zhou, Jing, additional, Wang, Yiqi, additional, Liu, Yonggang, additional, Guo, Chenfu, additional, Hu, Te, additional, Li, Congcong, additional, Liu, Leibo, additional, Zhang, Jin, additional, Wei, Shaojun, additional, Yang, Jun, additional, and Yin, Shouyi, additional

Published
2020
Full Text
View/download PDF

34. Electromagnetically Induced Transparency-Like Approach Based on Terahertz Metamaterials for Ultrasensitive Refractive Index Sensors.

Author

Cao, Pengfei, Li, Congcong, Li, Yuan, Wu, Yuyao, Chen, Xieyu, Wu, Tong, Zhang, Xueqian, Yuan, Mingrui, and Wang, Zelong

Abstract

In this paper, we design a sensor based on the electromagnetically induced transparency-like (EIT-like) approach in the terahertz range in order to realize a terahertz metamaterial device with an enhanced combination of ${Q}$ value and sensitivity. The Lorentz harmonic oscillator model and the asymmetry parameter are used for analyzing its internal physical mechanism. After processing the designed structure and performing experimental verification, a good consistency between the experiment and theoretical framework is found in terms of the position and strength of the resonance peak. Subsequently, glucose tablets of different concentrations are fabricated in order to verify the practical application of the designed sensor for distinguishing the presence of tablets with different concentrations. Finally, it is found through model optimization and sensor performance evaluation that our prototype exhibits a sensitivity value of about 330 GHz/RIU, a quality factor ${Q}$ of 24, while a recorded transparency peak above 80% is obtained. Therefore, as the metamaterial refractive index sensor presents an enhanced performance, it possesses great perspectives for various practical applications including the potential detection of different pharmaceuticals or biochemical products as well as the identification of various concentrations of the same sample. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

35. A Novel IGBT With Voltage-Clamping for Turn-on Overshoot Suppression Under Hard-Switching.

Author

Deng, Gaoqiang, Ma, Zhen, Luo, Xiaorong, Xie, Xintong, Li, Congcong, and Ng, Wai Tung

Subjects
INSULATED gate bipolar transistors, ELECTRON field emission, HIGH voltages
Abstract

A 600 V-rated insulated gate bipolar transistor (IGBT) with improved turn-on transient characteristics is proposed and investigated by simulation. The proposed IGBT features a built-in voltage-clamping structure. The floating ${p}$ -body adjacent to the trench gate is clamped at an appropriately high voltage before turn-on. This slows down the hole accumulation process in the floating p-body during turn-on transient, and thus the self-charging displacement current through the gate capacitance is reduced. The increasing rate for the gate voltage (dVGE/dt) is then reduced. As a result, the proposed IGBT achieves a 25% decrease in collector current ($I_C$) overshoot and 34% decrease in dIC/dt, when compared with the conventional IGBT with the same turn-on loss (EON), at IC = 250 A/cm2 under hard-switching. Its fabrication process is compatible with conventional trench IGBT. [ABSTRACT FROM AUTHOR]

Published
2021
Full Text
View/download PDF

36. TIMAQ: A Time-Domain Computing-in-Memory-Based Processor Using Predictable Decomposed Convolution for Arbitrary Quantized DNNs.

Author

Yang, Jianxun, Kong, Yuyao, Zhang, Zhao, Liu, Zhuangzhi, Zhou, Jing, Wang, Yiqi, Liu, Yonggang, Guo, Chenfu, Hu, Te, Li, Congcong, Liu, Leibo, Zhang, Jin, Wei, Shaojun, Yang, Jun, and Yin, Shouyi

Subjects
ENERGY consumption, TIME-domain analysis
Abstract

Energy-efficient processors are crucial for accelerating deep neural networks (DNNs) on edge devices with limited battery capacity. To reduce energy consumption, time-domain computing-in-memory (TD-CIM) is a splendid architecture, which consumes low computation and memory access energy due to low toggle rate of time-based signals and less data movements, respectively. When deploying DNNs in TD-CIMs, quantization is required, which has two types: uniform quantization (UQ) and nonuniform quantization (NUQ). To reach the same accuracy for one DNN, NUQ achieves smaller model size than UQ. Due to varying weight distributions across layers, mixed-precision quantization can further reduce model size, without degrading accuracy. However, previous TD-CIMs are inefficient for mixed-precision NUQ-DNNs due to their adopted bit-serial convolution increasing computation amount significantly. To address that, we propose a unique-weight convolution to accelerate mixed-precision NUQ-DNNs by a special kernel decomposition, reducing computation count remarkably. Based on that, we design a TD-CIM-based processor, TIMAQ, with three architectural techniques: 1) bit-cross-flipping-based kernel decomposer to reduce memory accesses and operations of decomposing kernels; 2) dual-mode-complementary predictor to remove redundant computations; and 3) activation-weight-adaptive pulse quantizer to decrease pulse quantization energy and error. Fabricated in 28-nm CMOS technology and tested on 1–8-b NUQ-DNNs, TIMAQ achieves 2.4–152.7-TOPS/W peak energy efficiency. [ABSTRACT FROM AUTHOR]

Published
2021
Full Text
View/download PDF

37. Experimental Realization of Ultralow ON-Resistance LDMOS With Optimized Layout.

Author

Wei, Jie, Ma, Zhen, Luo, Xiaorong, Li, Congcong, Song, Hua, Zhang, Sen, and Zhang, Bo

Subjects
BREAKDOWN voltage, ELECTRIC breakdown, METAL oxide semiconductor field-effect transistors, LOGIC circuits, DIODES
Abstract

To achieve lower specific ON-resistance (${R}_{{ \mathrm{\scriptscriptstyle ON},\text {SP}}}$), a novel accumulation lateral double-diffused MOSFET (ALDMOS) characterized by a convex-shaped field plate (CFP) structure is proposed and investigated experimentally. The CFP structure above the ${N}$ -drift region consists of two diodes, with its contact regions embedding into the drain and source region. Hence, the proposed device could shrink the drift length and utilize the drift region effectively to sustain the same level of breakdown voltage in the OFF-state. Furthermore, the CFP structure enables to form a continuous electron accumulation layer at the drift surface in the ON-state. Consequently, the proposed CFP ALDMOS could obtain an ultralow ${R}_{{ \mathrm{\scriptscriptstyle ON},\text {SP}}}$ and achieve a better tradeoff relationship between BV and ${R}_{{ \mathrm{\scriptscriptstyle ON},\text {SP}}}$. The experimental results show that the CFP ALDMOS realizes a BV of 448 V and ${R}_{{ \mathrm{\scriptscriptstyle ON},\text {SP}}}$ of 16.9 $\text{m}\Omega \cdot $ cm2 with an extremely high figure of merit of 11.87 MW/cm2, decreasing ${R}_{{ \mathrm{\scriptscriptstyle ON},\text {SP}}}$ by 55.6% compared with theoretical triple-REduced SURface Field (RESURF) LDMOS at the same BV. [ABSTRACT FROM AUTHOR]

Published
2021
Full Text
View/download PDF

38. The Detection of Low Frequency Oscillation Based on the Hilbert-Huang Transform Method

Author

Junqiang Tian, Kuan Li, Meng Liu, Li Congcong, Chao Yang, and Guohui Zhang

Subjects
Signal processing, Computer science, Noise (signal processing), Oscillation, Acoustics, 02 engineering and technology, Filter (signal processing), 01 natural sciences, Signal, Hilbert–Huang transform, Time–frequency analysis, 020303 mechanical engineering & transports, 0203 mechanical engineering, 0103 physical sciences, Low-frequency oscillation, 010301 acoustics
Abstract

A nonlinear and non-stationary signal processing method is proposed for the identification of low frequency oscillation in power system which is called Hilbert-Huang transform (HHT). In this paper, we first pretreat the noise and delay compensation of the oscillation signal, and then identify the modal parameters with the In this paper, we first pretreat the noise and delay compensation of the oscillation signal, and then identify the modal parameters with the HHT. Finally, the improved Prony algorithm identification results are compared with HHT. The simulation results show that the empirical mode decomposition can filter out the noise interference in the signal and lay the foundation for the accurate analysis of modal parameters. Moreover, the HHT method can effectively decompose the oscillating modes, and then accurately identify the modal parameters and obtain the time-frequency characteristics of the oscillation modes.

Published
2018

42. Study on fault diagnosis of electric energy data acquire system based on the BP neural network

Author

Zhu hongxia, Liu Xi Jian, Wang ZheLong, Li Congcong, and Du Yan

Subjects
Artificial neural network, Knowledge representation and reasoning, Java, Computer science, business.industry, Real-time computing, Point (geometry), Cloud computing, Algorithm design, Fault (power engineering), business, computer, computer.programming_language
Abstract

In this paper using BP neural network as core algorithm, Java and MATLAB as development tools, build a fault intelligent analysis system of electric energy data acquire network which have machine self-learning ability. This diagnosis system adopts the model of cloud computing in structure, artificial intelligence calculation in the cloud, friendly humancomputer interaction at the front-end. Lab simulation results showed that the system can analyze quickly and locate the fault point accurately. It can effectively reduce labor, time cost, improve work efficiency, has high practical value.

Published
2017

43. The research and rectification on the test of immunity to electromagnetic RF fields for electrical energy meter

Author

Wang Pingxin, Dai Yanjie, Yang Jie, Chen Zhiru, Zhang Zhi, Li Congcong, Dong Xianguang, Du Yan, and Li Lin

Subjects
Electromagnetic field, Electromagnetics, Rectification, Interference (communication), business.industry, Computer science, Electric potential energy, Electrical engineering, Metre, Radio frequency, business, Power (physics)
Abstract

With the wide application of electronic equipment and the popularization of electrical energy meter of state Grid, the electrical energy meter of site operation is often subject to various types of interference, the EMC of electrical energy meter become increasingly prominent. This methods and the problems of radio frequency electromagnetic field radiated immunity test are analyzed and discussed, the anti-jamming capability of electrical energy meter is improved by using the EMC technology. After rectification, the electrical energy meter passed the radio frequency electromagnetic field radiated immunity test, The measurement performance of the electrical energy meter is guaranteed, the legal rights and interests of both side of power supply and power demand are not damaged.

Published
2017

44. The research of performance testing about watt-hour meter plug

Author

Li Congcong, Wang Pingxin, Xu Ziqian, Dong Xianguang, and Zhang Zhi

Subjects
Watt, Smart meter, Computer science, 05 social sciences, 050301 education, 0102 computer and information sciences, 01 natural sciences, Automotive engineering, law.invention, Hardware_GENERAL, 010201 computation theory & mathematics, law, Metre, Metering mode, Spark plug, 0503 education
Abstract

As the important part of new low-voltage metering cabinet, the performance of watt-hour meter plug influences directly the convenience of operation, the stability of smart meter running and the reliability of metering cabinet. For lacking of the uniform technical specification and an effective, reliably and fast testing method for the performance of plug, this paper proposes a new testing method, it includes the important index of insertion and extraction power and temperature rise. The testing results shows that the method of this paper proposed can detect the mechanical and technical parameters accurately and conveniently, it provides reference for the stability and reliability of metering cabinet.

Published
2017

45. Experimental Study of 600 V Accumulation-Type Lateral Double-Diffused MOSFET With Ultra-Low On-Resistance.

Author

Deng, Gaoqiang, Luo, Xiaorong, Zhao, Zheyan, Wei, Jie, Cheng, Shikang, Li, Congcong, Ma, Zhen, Zhang, Bo, and Zhang, Sen

Subjects
BREAKDOWN voltage, METAL oxide semiconductor field-effect transistors, ELECTRIC fields, HIGH voltages, METAL semiconductor field-effect transistors
Abstract

A novel Lateral Double-Diffused MOSFET (LDMOSFET) with ultra-low resistance is investigated by experiments. The proposed LDMOSFET features an extended gate field plate consisting of two back-to-back diodes. In the ON state, the electron accumulation layer is formed below the gate field plate and greatly reduces the resistance of the drift region. In the OFF state, the field plate assists in depleting the drift region and modulates the electric field distribution of the drift region, which ensures high breakdown voltage (BV). The key fabrication process steps for the proposed device are given. The fabricated prototype device shows specific on-resistance (${R} _{{\text {on},\text {sp}}}$) of 70 $\text{m}\Omega \cdot $ cm2 and BV of 680 V, which is superior performance for reported 600 V-rated lateral devices. The ${R} _{{\text {on},\text {sp}}}$ of the proposed LDMOSFET is decreased by 50% compared with the LDMOSFET without accumulation effect. [ABSTRACT FROM AUTHOR]

Published
2020
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results