DOCSLIB.ORG

Fast Inference of Deep Neural Networks in Fpgas for Particle Physics (Arxiv:1804.06913)

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Fast Inference of Deep Neural Networks in Fpgas for Particle Physics (Arxiv:1804.06913) Fast Inference of Deep Neural Networks in FPGAs for Particle Physics (arXiv:1804.06913) Jennifer Ngadiuba, Maurizio Pierini [CERN] Javier Duarte, Sergo Jindariani, Ben Kreis, Ryan Rivera, Nhan Tran [Fermilab] Edward Kreinar [Hawkeye 360] Song Han, Phil Harris [MIT] Zhenbin Wu [University of Illinois at Chicago] Research Techniques Seminar Fermilab 4/24/2018 1 Outline • Introduction and Motivation • Machine Learning in HEP • FPGAs and High-Level Synthesis (HLS) • Industry Trends • HEP Latency Landscape • hls4ml: HLS for Machine Learning • Case Study and Design Exploration • Summary and Outlook • Cloud-scale Acceleration Javier Duarte I hls4ml 2 Introduction Javier Duarte I hls4ml 3 ReconstructionMachine chain:Learning Jet tagging in HEP Task to find the particle ID of a jet, e.g. b-quark • Learning optimized nonlinear functions of many inputs for … performingvertices difficult tasks from (real or simulated) data … Many successes in HEP: identificationData/MC of b-quarkuser jets, Higgs • Tag Info tagger corrections … candidates,Jet, particle energy regression, analysis selection, … particles CMS-PAS-BTV-15-001 s=13 TeV, 2016 1 CMS Simulation Preliminary Key features:tt events AK4jets (p > 30 GeV) T • Long lifetimeCSVv2 of heavy 10−1 flavour quarksDeepCSV cMVAv2 • Displacedmisid. probability tracks, … • Usage10−2 of ML standard for this problem udsg Neural network based on 10−3 • 11 c high-level features 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 b-jet efficiency Javier Duarte I hls4ml 4 Machine Learning in HEP • Learning optimized nonlinear functions of many inputs for performing difficult tasks from (real or simulated) data • Many successes in 11.1HEP: Significance identification of the signal and of its strengthb-quark jets, Higgs 41 candidates, particle energy regression, analysis selection, … arXiv:1407.0558 19.7 fb-1 (8 TeV) + 5.1 fb-1 (7 TeV) ×1043 CMS 3.5 S/(S+B) weighted sum H → γγ Data 3 ML algorithms used in every S+B fits (weighted sum) 2.5 B component aspect of Higgs discovery: ±1σ energy regression 2 ±2σ S/B discrimination, 1.5 1 +0.26 µ = 1.14 −0.23 … 0.5 mH = 124.70 ± 0.34 GeV 0 S/(S+B) weighted events / GeV 200 Typically applied offline, B component subtracted not online (trigger-level) 100 0 -100 110 115 120 125 130 135 140 145 150 Javier Duarte I hls4ml 5 mγγ (GeV) Figure 19: Diphoton mass spectrum weighted by the ratio S/(S + B) in each event class, to- gether with the background subtracted weighted mass spectrum. Table 5: Values of the best-fit signal strength, µˆ, when mH is treated as an unconstrained pa- rameter, for the 7 TeV, 8 TeV, and combined datasets. The corresponding best-fit value of mH, mH, is also given. µˆ mH (GeV) b +0.62 7 TeV 2.22 0.55 124.2 +−0.26 b 8 TeV 0.90 0.23 124.9 +−0.26 Combined 1.14 0.23 124.7 − section times the relevant branching fractions, relative to the SM expectation. In Fig. 20 the combined best-fit signal strength, µˆ, is shown as a function of the Higgs boson mass hypothesis, both for the standard analysis (left) and for the cut-based analysis (right). The two analyses agree well across the entire mass range. In addition to the signal around 125 GeV, both analyses see a small upward fluctuation at 150 GeV, which is found to have a maximum local significance of just over 2 s at mH = 151 GeV—slightly beyond the mass range of our analysis. The best-fit signal strength for the main analysis, when the value of mH is treated as an un- +0.26 constrained parameter in the fit, is µˆ = 1.14 0.23, with the corresponding best-fit mass being − mH = 124.7 GeV. The expected uncertainties in the best-fit signal strength, at this mass, are +0.24 and 0.22. The values of the best-fit signal strength, derived separately for the 7 and − 8b TeV datasets, are listed in Table 5. For the cut-based analysis the corresponding value is +0.29 µˆ = 1.29 0.26 at mH = 124.6 GeV, and for the sideband background model analysis the value − +0.26 measured is µˆ = 1.06 0.23 at mH = 124.7 GeV. These values are shown in Table 6 together with the expected uncertainty,b − and the corresponding values for the main analysis. The uncertainty in the signalb strength may be separated into statistical and systematic con- tributions, with the latter further divided into those having, or not, a theoretical origin: µˆ = CMS Trigger 1 kHz 100 kHz 1 MB/evt High-Level 40 MHz Trigger L1 Trigger • Level-1 Trigger (hardware) • High-Level Trigger (software) • 99.75% rejected • 99% rejected • decision in ~4 μs • decision in ~100s ms • After trigger, 99.99975% of events are gone forever Javier Duarte I hls4ml 6 HEP Latency Landscape CMS: L1 Trigger HLT Offline 1 ns 1 us 1 ms 1 s FPGAs CPUs Javier Duarte I hls4ml 7 MACHINE LEARNING & FPGAS 7 Machine learning algorithms are ubiquitous in HEP FPGA usage broad across HEP experiments Field-Programmable Gate Array Centered on DAQ and triggerALL FPGA development ARCHITECTURE 16 • Flexible: re-programmable Some early adaptions ofO(50-100) ML techniques optical in triggerFPGA [1] interconnects betweentransceivers “programmable hardware” FPGA developmentconfigurable becominglogic blocksrunning atmore and accessible ~O(15) Gbs High Levelembedded Synthesis, components OpenCL • LUTs (logic), What are FPGAs?FPGA Flip-Flops (registers), “programmable hardware” FPGA diagram Field Programmable Gate Arrays are reprogrammable FPGA interestDSPs in (arithmetic),industry is growingintegrated circuits Contain array of logic blocks used to configure low level Block RAMs (memory) operations (bit masking,DSPs shifting, (multiply-accumulate,addition) etc.) Programmable hardware with structures Flip Flops (registers/distributed memory) Logic blocks wired together through routing channels High Throughput: O(100)Typical modern FPGA: LUTs (logic) that maps• nicely onto ML architecturesSupport highly parallel and pipelinedBlock algorithm RAMs (memories) (Kintex ultrascale+) implementations with guaranteed latency optical transceivers 1.3Mrunning FFs at 700k LUTs O(15) Gbs 5500 DSPs Typical modern FPGADSPs (multiply-accumulate,LogicLogic block cell etc.) 2200 BRAMs (Kintex UltraScale):Flip Flops (registers/distributed memory) • Massively parallel 1.3 M FFs LUTs (logic) Block RAMsLook-up (memories) Flip-flop 700k LUTs table • Low power (relative to CPU/ 5500 DSPs 2200 BRAMs 9 GPU) 25.04.2018 Jennifer Ngadiuba - hls4ml: deep neural networks in FPGAs [1] Carnes et al., https://indico.cern.ch/event/567550/contributions/2629686/ Javier Duarte I hls4ml 8 CPUs, GPUs, FPGAs, and ASICs How are they different? • FPGAs are the middle ground of latency, energy AeRCHITECTURESfficiency, and flexibility 41 ASICs FPGAs GPUs CPUs Source: Bob Broderson, Berkeley Wireless group Javier* GPUs Duarte still Ibest hls4ml option for training 9 * FPGAs generally much more power efficient High-Level Synthesis • FPGA development is becoming more accessible, with tools like High-Level Synthesis: untimed C code with additional Motivation for High-Level Synthesis (HLS) directives that is synthesized into RTL Verilog/VHDL Motivationmodule dut(rst, clk for, q); High-Level Synthesis (HLS) input rst; uint8 dut() { input clk; static uint8 c; moduleoutputdutq;(rst , clk, q); c+=1; inputregrst[7:0]; c; vs. } uint8 dut() { clk; inputalways @ (posedge clk) static uint8 c; Untimed C code outputbeginq; c+=1; reg [7:0]if (rst c;== 1b’1) begin vs. } c <= 8'b00000000; HLS alwaysend @ (posedge clk) Untimed C code beginelse begin rst 1 c <= c + 1; if (rst == 1b’1) begin end + q c <= 8'b00000000; 1 0 HLS 0 c 8 end assign q = c; elseendmodule begin rst 1 clk c <= c RTL+ 1; Verilog Javier Duarteend I hls4ml 10 + q An1 0 8-bit counter 0 c 8 4 assign q = c; endmodule clk RTL Verilog An 8-bit counter 4 Industry Trends • Industry is moving toward custom hardware and FPGAs to quickly apply ML algorithms • Already being used as backend accelerators for Bing web searches, Siri queries, and more… Microsoft Apple Google Intel Javier Duarte I hls4ml 11 hls4ml 100 kHz 40 MHz L1 Trigger • hls4ml: neural network translation library for HLS • Support for common ML workflows and architectures • Tunable configuration for different use cases • Focus on L1 trigger as 1st application • What can we do in < μs on one FPGA? Javier Duarte I hls4ml 12 Case Study and Design Exploration Javier Duarte I hls4ml 13 Case Study: Jet Substructure • Illustrative example: not necessarily the most realistic for L1 today, but lessons are generic Javier Duarte I hls4ml 14 Jet Substructure Inputs ECFs Observables mass mmMDT β=1,2 N2 β=1,2 M2 β=0,1,2 C1 β=1,2 C2 β=1,2 D2 ↵,β = 1,1 , 1,2 multiplicity ( ) ( ) ( ) D2 z log z Multiplicity Õ Table 1: A summary of the observables used in the analysis. Groomed mass separates top, W/Z, and quark/gluon this study [51–54]. A brief description of• each of these variables is presented in Ref. [55]. These are used as expert-level inputs to a neural network• top/gluon classifier which have is greater near optimal multiplicity3. than W/Z/quark • ECF N2β=1 separates 2 and 3-prong jets (W/Z/top) from 1-prong jets (quark/gluon) Benchmark networks and floating point performance We train a neural network for the classification task of q, g, W, Z, and t discrimination. The data are Javier Duarte I hls4ml 15 randomly split into training (60%), validation (20%), and testing (20%) datasets. The input features are standardized by removing the mean and scaling to unit variance. The architecture, illustrated in Fig. 4 (left), is a fully-connected neural network with three hidden layers.
Load more

Recommended publications
  • Intelligent Recognition Method of Low-Altitude Squint Optical Ship Target Fused with Simulation Samples
    remote sensing Article Intelligent Recognition Method of Low-Altitude Squint Optical Ship Target Fused with Simulation Samples Bo Liu 1,2 , Qi Xiao 1,2, Yuhao Zhang 1,2, Wei Ni 1, Zhen Yang 1 and Ligang Li 1,* 1 National Space Science Center, Key Laboratory of Electronics and Information Technology for Space System, Chinese Academy of Sciences, Beijing 100190, China; [email protected] (B.L.); [email protected] (Q.X.); [email protected] (Y.Z.); [email protected] (W.N.); [email protected] (Z.Y.) 2 School of Computer Science and Technology, University of Chinese Academy of Sciences, Beijing 100049, China * Correspondence: [email protected]; Tel.: +86-1312-152-1820 Abstract: To address the problem of intelligent recognition of optical ship targets under low-altitude squint detection, we propose an intelligent recognition method based on simulation samples. This method comprehensively considers geometric and spectral characteristics of ship targets and ocean background and performs full link modeling combined with the squint detection atmospheric transmission model. It also generates and expands squint multi-angle imaging simulation samples of ship targets in the visible light band using the expanded sample type to perform feature analysis and modification on SqueezeNet. Shallow and deeper features are combined to improve the accuracy of feature recognition. The experimental results demonstrate that using simulation samples to expand the training set can improve the performance of the traditional k-nearest neighbors algorithm and modified SqueezeNet. For the classification of specific ship target types, a mixed-scene dataset Citation: Liu, B.; Xiao, Q.; Zhang, Y.; expanded with simulation samples was used for training.
    [Show full text]
  • NXP Eiq Machine Learning Software Development Environment For
    AN12580 NXP eIQ™ Machine Learning Software Development Environment for QorIQ Layerscape Applications Processors Rev. 3 — 12/2020 Application Note Contents 1 Introduction 1 Introduction......................................1 Machine Learning (ML) is a computer science domain that has its roots in 2 NXP eIQ software introduction........1 3 Building eIQ Components............... 2 the 1960s. ML provides algorithms capable of finding patterns and rules in 4 OpenCV getting started guide.........3 data. ML is a category of algorithm that allows software applications to become 5 Arm Compute Library getting started more accurate in predicting outcomes without being explicitly programmed. guide................................................7 The basic premise of ML is to build algorithms that can receive input data and 6 TensorFlow Lite getting started guide use statistical analysis to predict an output while updating output as new data ........................................................ 8 becomes available. 7 Arm NN getting started guide........10 8 ONNX Runtime getting started guide In 2010, the so-called deep learning started. It is a fast-growing subdomain ...................................................... 16 of ML, based on Neural Networks (NN). Inspired by the human brain, 9 PyTorch getting started guide....... 17 deep learning achieved state-of-the-art results in various tasks; for example, A Revision history.............................17 Computer Vision (CV) and Natural Language Processing (NLP). Neural networks are capable of learning complex patterns from millions of examples. A huge adaptation is expected in the embedded world, where NXP is the leader. NXP created eIQ machine learning software for QorIQ Layerscape applications processors, a set of ML tools which allows developing and deploying ML applications on the QorIQ Layerscape family of devices.
    [Show full text]
  • Comparison and Benchmarking of AI Models and Frameworks on Mobile Devices
    Comparison and Benchmarking of AI Models and Frameworks on Mobile Devices 1st Chunjie Luo 2nd Xiwen He 3nd Jianfeng Zhan Institute of Computing Technology Institute of Computing Technology Institute of Computing Technology Chinese Academy of Sciences Chinese Academy of Sciences Chinese Academy of Sciences BenchCouncil BenchCouncil BenchCouncil Beijing, China Beijing, China Beijing, China [email protected] [email protected] [email protected] 4nd Lei Wang 5nd Wanling Gao 6nd Jiahui Dai Institute of Computing Technology Institute of Computing Technology Beijing Academy of Frontier Chinese Academy of Sciences Chinese Academy of Sciences Sciences and Technology BenchCouncil BenchCouncil Beijing, China Beijing, China Beijing, China [email protected] wanglei [email protected] [email protected] Abstract—Due to increasing amounts of data and compute inference on the edge devices can 1) reduce the latency, 2) resources, deep learning achieves many successes in various protect the privacy, 3) consume the power [2]. domains. The application of deep learning on the mobile and em- To make the inference on the edge devices more efficient, bedded devices is taken more and more attentions, benchmarking and ranking the AI abilities of mobile and embedded devices the neural networks are designed more light-weight by using becomes an urgent problem to be solved. Considering the model simpler architecture, or by quantizing, pruning and compress- diversity and framework diversity, we propose a benchmark suite, ing the networks. Different networks present different trade- AIoTBench, which focuses on the evaluation of the inference offs between accuracy and computational complexity. These abilities of mobile and embedded devices.
    [Show full text]
  • An FPGA-Accelerated Embedded Convolutional Neural Network
    Master Thesis Report ZynqNet: An FPGA-Accelerated Embedded Convolutional Neural Network (edit) (edit) 1000ch 1000ch FPGA 1000ch Network Analysis Network Analysis 2x512 > 1024 2x512 > 1024 David Gschwend [email protected] SqueezeNet v1.1 b2a ext7 conv10 2x416 > SqueezeNet SqueezeNet v1.1 b2a ext7 conv10 2x416 > SqueezeNet arXiv:2005.06892v1 [cs.CV] 14 May 2020 Supervisors: Emanuel Schmid Felix Eberli Professor: Prof. Dr. Anton Gunzinger August 2016, ETH Zürich, Department of Information Technology and Electrical Engineering Abstract Image Understanding is becoming a vital feature in ever more applications ranging from medical diagnostics to autonomous vehicles. Many applications demand for embedded solutions that integrate into existing systems with tight real-time and power constraints. Convolutional Neural Networks (CNNs) presently achieve record-breaking accuracies in all image understanding benchmarks, but have a very high computational complexity. Embedded CNNs thus call for small and efficient, yet very powerful computing platforms. This master thesis explores the potential of FPGA-based CNN acceleration and demonstrates a fully functional proof-of-concept CNN implementation on a Zynq System-on-Chip. The ZynqNet Embedded CNN is designed for image classification on ImageNet and consists of ZynqNet CNN, an optimized and customized CNN topology, and the ZynqNet FPGA Accelerator, an FPGA-based architecture for its evaluation. ZynqNet CNN is a highly efficient CNN topology. Detailed analysis and optimization of prior topologies using the custom-designed Netscope CNN Analyzer have enabled a CNN with 84.5 % top-5 accuracy at a computational complexity of only 530 million multiply- accumulate operations. The topology is highly regular and consists exclusively of convolu- tional layers, ReLU nonlinearities and one global pooling layer.
    [Show full text]
  • Compositional Falsification of Cyber-Physical Systems with Machine Learning Components
    Compositional Falsification of Cyber-Physical Systems with Machine Learning Components Tommaso Dreossi Alexandre Donze Sanjit A. Seshia Electrical Engineering and Computer Sciences University of California at Berkeley Technical Report No. UCB/EECS-2017-165 http://www2.eecs.berkeley.edu/Pubs/TechRpts/2017/EECS-2017-165.html November 26, 2017 Copyright © 2017, by the author(s). All rights reserved. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission. Acknowledgement This technical report is an extended version of an article that appeared at NFM 2017 and is under submission to the Journal of Automated Reasoning (JAR). Noname manuscript No. (will be inserted by the editor) Compositional Falsification of Cyber-Physical Systems with Machine Learning Components Tommaso Dreossi · Alexandre Donz´e · Sanjit A. Seshia the date of receipt and acceptance should be inserted later Abstract Cyber-physical systems (CPS), such as automotive systems, are starting to include sophisticated machine learning (ML) components. Their correctness, therefore, depends on properties of the inner ML modules. While learning algorithms aim to generalize from examples, they are only as good as the examples provided, and recent efforts have shown that they can pro- duce inconsistent output under small adversarial perturbations. This raises the question: can the output from learning components lead to a failure of the entire CPS? In this work, we address this question by formulating it as a problem of falsifying signal temporal logic (STL) specifications for CPS with ML components.
    [Show full text]
  • Firenet Architectures 510X Smaller Models
    [email protected] Architecting new deep neural networks for embedded applications Forrest Iandola 1 [email protected] Machine Learning in 2012 Object Detection Sentiment Analysis Deformable Parts Model LDA Text Computer Analysis Vision Semantic Segmentation Word Prediction Audio segDPM Linear Interpolation Analysis + N-Gram Speech Recognition Audio Concept Image Classification Recognition Hidden Markov Feature Engineering Model + SVMs i-VeCtor + HMM We have 10 years of experienCe in a broad variety of ML approaChes … [1] B. Catanzaro, N. Sundaram, K. Keutzer. Fast support veCtor maChine training and ClassifiCation on graphiCs proCessors. International ConferenCe on MaChine Learning (ICML), 2008. [2] Y. Yi, C.Y. Lai, S. Petrov, K. Keutzer. EffiCient parallel CKY parsing on GPUs. International ConferenCe on Parsing TeChnologies, 2011. [3] K. You, J. Chong, Y. Yi, E. Gonina, C.J. Hughes, Y. Chen, K. Keutzer. Parallel scalability in speeCh reCognition. IEEE Signal Processing Magazine, 2009. [4] F. Iandola, M. Moskewicz, K. Keutzer. libHOG: Energy-EffiCient Histogram of Oriented Gradient Computation. ITSC, 2015. [5] N. Zhang, R. Farrell, F. Iandola, and T. Darrell. Deformable Part DesCriptors for Fine-grained ReCognition and Attribute PrediCtion. ICCV, 2013. [6] M. Kamali, I. Omer, F. Iandola, E. Ofek, and J.C. Hart. Linear Clutter Removal from Urban Panoramas International Symposium on Visual Computing. ISVC, 2011. 2 [email protected] By 2016, Deep Neural Networks Give Superior Solutions in Many Areas Sentiment Analysis ObjeCt DeteCtion 3-layer RNN 16-layer DCNN Text CNN/ Computer Analysis DNN Vision Semantic Segmentation Word PrediCtion word2veC NN 19-layer FCN Audio Speech ReCognition Analysis Audio Concept Image ClassifiCation ReCognition LSTM NN GoogLeNet-v3 DCNN 4-layer DNN Finding the "right" DNN arChiteCture is replaCing broad algorithmiC exploration for many problems.
    [Show full text]
  • Serving Deep Learning Models in a Serverless Platform
    Serving deep learning models in a serverless platform Vatche Ishakian Vinod Muthusamy Aleksander Slominski [email protected] [email protected] [email protected] Bentley University IBM T.J. Watson Research Center IBM T.J. Watson Research Center Abstract—Serverless computing has emerged as a compelling network for inferencing (a.k.a serving), on the other hand, paradigm for the development and deployment of a wide range of requires no training data and less computational power than event based cloud applications. At the same time, cloud providers training them. and enterprise companies are heavily adopting machine learning and Artificial Intelligence to either differentiate themselves, or In this work, we attempt to evaluate the suitability of provide their customers with value added services. In this work serverless computing to run deep learning inferencing tasks. we evaluate the suitability of a serverless computing environ- We measure the performance as seen by the user, and the cost ment for the inferencing of large neural network models. Our of running three different MXNet [7] trained deep learning experimental evaluations are executed on the AWS Lambda models on the AWS Lambda serverless computing platform. environment using the MxNet deep learning framework. Our experimental results show that while the inferencing latency can Our preliminary set of experimental results show that a server- be within an acceptable range, longer delays due to cold starts less platform is suitable for deploying deep learning prediction can skew the latency distribution and hence risk violating more services, but the delays occured as a result of running the first stringent SLAs.
    [Show full text]
  • High Performance Squeezenext for CIFAR-10
    High Performance SqueezeNext for CIFAR-10 Jayan Kant Duggal and Mohamed El-Sharkawy Electrical and Computer Engineering IoT Collaboratory, Purdue School of Engineering and Technology, IUPUI [email protected], [email protected] Abstract—CNNs is the foundation for deep learning and deployability of these models on real-time hardware systems computer vision domain enabling applications such as such as advanced driver-assistance systems (ADAS), drones, autonomous driving, face recognition, automatic radiology edge devices, robotics, UAVs, and all other real-time image reading, etc. But, CNN is a algorithm which is memory and computationally intensive. DSE of neural networks and applications that require low-cost and low-power. In Section compression techniques have made convolution neural networks 2 of this paper, the related architectures, SqueezeNet and memory and computationally efficient. It improved the CNN SqueezeNext architectures along with their observed architectures and made it more suitable to implement on problems were reviewed. Subsequently, in Sections 3, the real-time embedded systems. This paper proposes an efficient general methods to improve CNN performance such as and a compact CNN to ameliorate the performance of existing CNN architectures. The intuition behind this proposed architecture tuning, different learning rate methods, save and architecture is to supplant convolution layers with a more load checkpoint method, different types of optimizers and sophisticated block module and to develop a compact different activation functions were discussed. This section architecture with a competitive accuracy. Further, explores the reflects the DSE of DNNs which is essential in building or bottleneck module and squeezenext basic block structure. The improving any CNN/DNN.
    [Show full text]
  • Investigations of Object Detection in Images/Videos Using Various Deep Learning Techniques and Embedded Platforms—A Comprehensive Review
    applied sciences Review Investigations of Object Detection in Images/Videos Using Various Deep Learning Techniques and Embedded Platforms—A Comprehensive Review Chinthakindi Balaram Murthy 1,†, Mohammad Farukh Hashmi 1,† , Neeraj Dhanraj Bokde 2,† and Zong Woo Geem 3,* 1 Department of Electronics and Communication Engineering, National Institute of Technology, Warangal 506004, India; [email protected] (C.B.M.); [email protected] (M.F.H.) 2 Department of Engineering—Renewable Energy and Thermodynamics, Aarhus University, 8000 Aarhus, Denmark; [email protected] 3 Department of Energy IT, Gachon University, Seongnam 13120, Korea * Correspondence: [email protected]; Tel.: +82-31-750-5586 † These authors contributed equally to this work. Received: 15 April 2020; Accepted: 2 May 2020; Published: 8 May 2020 Abstract: In recent years there has been remarkable progress in one computer vision application area: object detection. One of the most challenging and fundamental problems in object detection is locating a specific object from the multiple objects present in a scene. Earlier traditional detection methods were used for detecting the objects with the introduction of convolutional neural networks. From 2012 onward, deep learning-based techniques were used for feature extraction, and that led to remarkable breakthroughs in this area. This paper shows a detailed survey on recent advancements and achievements in object detection using various deep learning techniques. Several topics have been included, such as Viola–Jones (VJ), histogram of oriented gradient (HOG), one-shot and two-shot detectors, benchmark datasets, evaluation metrics, speed-up techniques, and current state-of-art object detectors. Detailed discussions on some important applications in object detection areas, including pedestrian detection, crowd detection, and real-time object detection on Gpu-based embedded systems have been presented.
    [Show full text]
  • Real-Time Vehicle Make and Model Recognition with the Residual Squeezenet Architecture
    sensors Article Real-Time Vehicle Make and Model Recognition with the Residual SqueezeNet Architecture Hyo Jong Lee 1,* , Ihsan Ullah 2, Weiguo Wan 1, Yongbin Gao 3 and Zhijun Fang 3 1 Division of Computer Science and Engineering, CAIIT, Chonbuk National University, Jeonju 54896, Korea; [email protected] 2 Department of Robotics Engineering, Daegu Gyeongbuk Institute of Science and Technology, Daegu 42988, Korea; [email protected] 3 School of Electrics and Electronic Engineering, Shanghai University of Engineering Science, Shanghai 201620, China; [email protected] (Y.G.); [email protected] (Z.F.) * Correspondence: [email protected]; Tel.: +82-63-270-2407 Received: 20 December 2018; Accepted: 21 February 2019; Published: 26 February 2019 Abstract: Make and model recognition (MMR) of vehicles plays an important role in automatic vision-based systems. This paper proposes a novel deep learning approach for MMR using the SqueezeNet architecture. The frontal views of vehicle images are first extracted and fed into a deep network for training and testing. The SqueezeNet architecture with bypass connections between the Fire modules, a variant of the vanilla SqueezeNet, is employed for this study, which makes our MMR system more efficient. The experimental results on our collected large-scale vehicle datasets indicate that the proposed model achieves 96.3% recognition rate at the rank-1 level with an economical time slice of 108.8 ms. For inference tasks, the deployed deep model requires less than 5 MB of space and thus has a great viability in real-time applications. Keywords: vehicle make recognition; deep learning; residual SqueezeNet 1. Introduction In recent years, a plethora of innovative technologies and solutions are bringing intelligent transportation systems (ITSs) closer to reality.
    [Show full text]
  • Helping the Visually Impaired See Via Image Multi-Labeling Based on Squeezenet CNN
    applied sciences Article Helping the Visually Impaired See via Image Multi-labeling Based on SqueezeNet CNN Haikel Alhichri * , Yakoub Bazi , Naif Alajlan and Bilel Bin Jdira Advanced Lab for Intelligent Systems research (ALISR), Computer Engineering, College of Comp. and Info. Sciences, King Saud University, Riyadh 11543, Saudi Arabia; [email protected] (Y.B.); [email protected] (N.A.); [email protected] (B.B.J.) * Correspondence: [email protected]; Tel.: +966-11-469-6294 Received: 28 September 2019; Accepted: 30 October 2019; Published: 1 November 2019 Abstract: This work presents a deep learning method for scene description. (1) Background: This method is part of a larger system, called BlindSys, that assists the visually impaired in an indoor environment. The method detects the presence of certain objects, regardless of their position in the scene. This problem is also known as image multi-labeling. (2) Methods: Our proposed deep learning solution is based on a light-weight pre-trained CNN called SqueezeNet. We improved the SqueezeNet architecture by resetting the last convolutional layer to free weights, replacing its activation function from a rectified linear unit (ReLU) to a LeakyReLU, and adding a BatchNormalization layer thereafter. We also replaced the activation functions at the output layer from softmax to linear functions. These adjustments make up the main contributions in this work. (3) Results: The proposed solution is tested on four image multi-labeling datasets representing different indoor environments. It has achieved results better than state-of-the-art solutions both in terms of accuracy and processing time. (4) Conclusions: The proposed deep CNN is an effective solution for predicting the presence of objects in a scene and can be successfully used as a module within BlindSys.
    [Show full text]
  • Deep Convolutional Neural Networks for Tiny Imagenet Classification
    Deep Convolutional Neural Networks for Tiny ImageNet Classification Hujia Yu Stanford University [email protected] Abstract cently. Since the 2012 ImageNet competition winning en- try by Krizhevsky et al.[1], their network AlexNet has been In this project, I approached image classification prob- applied to variety of computer vision tasks, for example to lem by implementing and training from scratch three state- object detection. Successful models like AlexNet has great of-art model structures, AlexNet, GoogLeNet, and ResNet influence on research that focused on deeper and better- on the TinyImageNet dataset. After fine-tuning and evaluat- performing network structures. In 2014, the quality of net- ing the performance on all three models, I used my best per- work architecture significantly improved by using deeper forming model, which is ResNet-20 with 55.4% validation and wider networks. VGGNet[7] and GoogLeNet[5] both accuracy, on the test set, and achieved test accuracy of 45% yielded high performance in ILSVRC 2014 classification (error rate of 55% ). During this process, I found out that challenge. This paper uses dataset from Tiny ImageNet most of the models trained from scratch tend to plateau at Challenge, which is a simplified version of Large Scale Vi- around or below 50% validation accuracy. It turns out that sual Recognition Challenge 2016(ILSVRC2016)[10]. The training state-of-art model architectures from scratch on re- details of the dataset is described in the section below. In duced dimension of image datasets do not have as good per- this paper, I trained from scratch the reduced capacity mod- formances as the original models on the Large Scale Im- els of GoogLeNet, AlexNet, and ResNet[6] on the dataset, ageNet Challenge.
    [Show full text]