From the Publisher: 
The accessible presentation of this book gives both a general view of the entire computer vision enterprise and also offers sufficient detail to be able to build useful applications. Users learn techniques that have proven to be useful by first-hand experience and a wide range of mathematical methods. A CD-ROM with every copy of the text contains source code for programming practice, color images, and illustrative movies. Comprehensive and up-to-date, this book includes essential topics that either reflect practical significance or are of theoretical importance. Topics are discussed in substantial and increasing depth. Application surveys describe numerous important application areas such as image based rendering and digital libraries. Many important algorithms broken down and illustrated in pseudo code. Appropriate for use by engineers as a comprehensive reference to the computer vision enterprise.

Computer vision : a modern approach = 计算机视觉 : 一种现代的方法

Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU)---deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs (caches, out-of-order execution, multithreading, multiprocessing, prefetching, ...) that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X - 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X - 80X higher. Moreover, using the GPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.

In-Datacenter Performance Analysis of a Tensor Processing Unit

Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU) --- deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X -- 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X -- 80X higher. Moreover, using the CPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.

/pdf/in-datacenter-performance-analysis-of-a-tensor-processing-4pmp83zqw4.pdf

State-of-the-art deep neural networks (DNNs) have hundreds of millions of connections and are both computationally and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources and power budgets. While custom hardware helps the computation, fetching weights from DRAM is two orders of magnitude more expensive than ALU operations, and dominates the required power.Previously proposed 'Deep Compression' makes it possible to fit large DNNs (AlexNet and VGGNet) fully in on-chip SRAM. This compression is achieved by pruning the redundant connections and having multiple connections share the same weight. We propose an energy efficient inference engine (EIE) that performs inference on this compressed network model and accelerates the resulting sparse matrix-vector multiplication with weight sharing. Going from DRAM to SRAM gives EIE 120× energy saving; Exploiting sparsity saves 10×; Weight sharing gives 8×; Skipping zero activations from ReLU saves another 3×. Evaluated on nine DNN benchmarks, EIE is 189× and 13× faster when compared to CPU and GPU implementations of the same DNN without compression. EIE has a processing power of 102 GOPS working directly on a compressed network, corresponding to 3 TOPS on an uncompressed network, and processes FC layers of AlexNet at 1.88×104 frames/sec with a power dissipation of only 600mW. It is 24,000× and 3,400× more energy efficient than a CPU and GPU respectively. Compared with DaDianNao, EIE has 2.9×, 19× and 3× better throughput, energy efficiency and area efficiency.

EIE: efficient inference engine on compressed deep neural network

Deep neural networks (DNNs) are currently widely used for many artificial intelligence (AI) applications including computer vision, speech recognition, and robotics. While DNNs deliver state-of-the-art accuracy on many AI tasks, it comes at the cost of high computational complexity. Accordingly, techniques that enable efficient processing of DNNs to improve energy efficiency and throughput without sacrificing application accuracy or increasing hardware cost are critical to the wide deployment of DNNs in AI systems. This article aims to provide a comprehensive tutorial and survey about the recent advances toward the goal of enabling efficient processing of DNNs. Specifically, it will provide an overview of DNNs, discuss various hardware platforms and architectures that support DNNs, and highlight key trends in reducing the computation cost of DNNs either solely via hardware design changes or via joint hardware design and DNN algorithm changes. It will also summarize various development resources that enable researchers and practitioners to quickly get started in this field, and highlight important benchmarking metrics and design considerations that should be used for evaluating the rapidly growing number of DNN hardware designs, optionally including algorithmic codesigns, being proposed in academia and industry. The reader will take away the following concepts from this article: understand the key design considerations for DNNs; be able to evaluate different DNN hardware implementations with benchmarks and comparison metrics; understand the tradeoffs between various hardware architectures and platforms; be able to evaluate the utility of various DNN design techniques for efficient processing; and understand recent implementation trends and opportunities.

Efficient Processing of Deep Neural Networks: A Tutorial and Survey

Convolutional neural network (CNN) has been widely employed for image recognition because it can achieve high accuracy by emulating behavior of optic nerves in living creatures. Recently, rapid growth of modern applications based on deep learning algorithms has further improved research and implementations. Especially, various accelerators for deep CNN have been proposed based on FPGA platform because it has advantages of high performance, reconfigurability, and fast development round, etc. Although current FPGA accelerators have demonstrated better performance over generic processors, the accelerator design space has not been well exploited. One critical problem is that the computation throughput may not well match the memory bandwidth provided an FPGA platform. Consequently, existing approaches cannot achieve best performance due to under-utilization of either logic resource or memory bandwidth. At the same time, the increasing complexity and scalability of deep learning applications aggravate this problem. In order to overcome this problem, we propose an analytical design scheme using the roofline model. For any solution of a CNN design, we quantitatively analyze its computing throughput and required memory bandwidth using various optimization techniques, such as loop tiling and transformation. Then, with the help of rooine model, we can identify the solution with best performance and lowest FPGA resource requirement. As a case study, we implement a CNN accelerator on a VC707 FPGA board and compare it to previous approaches. Our implementation achieves a peak performance of 61.62 GFLOPS under 100MHz working frequency, which outperform previous approaches significantly.

Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks

Escalating system-on-chip design complexity is pushing the design community to raise the level of abstraction beyond register transfer level. Despite the unsuccessful adoptions of early generations of commercial high-level synthesis (HLS) systems, we believe that the tipping point for transitioning to HLS msystem-on-chip design complexityethodology is happening now, especially for field-programmable gate array (FPGA) designs. The latest generation of HLS tools has made significant progress in providing wide language coverage and robust compilation technology, platform-based modeling, advancement in core HLS algorithms, and a domain-specific approach. In this paper, we use AutoESL's AutoPilot HLS tool coupled with domain-specific system-level implementation platforms developed by Xilinx as an example to demonstrate the effectiveness of state-of-art C-to-FPGA synthesis solutions targeting multiple application domains. Complex industrial designs targeting Xilinx FPGAs are also presented as case studies, including comparison of HLS solutions versus optimized manual designs. In particular, the experiment on a sphere decoder shows that the HLS solution can achieve an 11-31% reduction in FPGA resource usage with improved design productivity compared to hand-coded design.

/pdf/high-level-synthesis-for-fpgas-from-prototyping-to-4vlvkebjds.pdf

High-Level Synthesis for FPGAs: From Prototyping to Deployment

The field programmable gate-array (FPGA) has become an important technology in VLSI ASIC designs. In the past few years, a number of heuristic algorithms have been proposed for technology mapping in lookup-table (LUT) based FPGA designs, but none of them guarantees optimal solutions for general Boolean networks and little is known about how far their solutions are away from the optimal ones. This paper presents a theoretical breakthrough which shows that the LUT-based FPGA technology mapping problem for depth minimization can be solved optimally in polynomial time. A key step in our algorithm is to compute a minimum height K-feasible cut in a network, which is solved optimally in polynomial time based on network flow computation. Our algorithm also effectively minimizes the number of LUT's by maximizing the volume of each cut and by several post-processing operations. Based on these results, we have implemented an LUT-based FPGA mapping package called FlowMap. We have tested FlowMap on a large set of benchmark examples and compared it with other LUT-based FPGA mapping algorithms for delay optimization, including Chortle-d, MIS-pga-delay, and DAG-Map. FlowMap reduces the LUT network depth by up to 7% and reduces the number of LUT's by up to 50% compared to the three previous methods. >

/pdf/flowmap-an-optimal-technology-mapping-algorithm-for-delay-59gk35bj0r.pdf

FlowMap: an optimal technology mapping algorithm for delay optimization in lookup-table based FPGA designs

As the technology progresses, interconnect delays have become bottlenecks of chip performance. 3D integrated circuits are proposed as one way to address this problem. However, thermal problem is a critical challenge for 3D IC circuit design. We propose a thermal-driven 3D floorplanning algorithm. Our contributions include: (1) a new 3D floorplan representation, CBA and new interlayer local operations to more efficiently exploit the solution space; (2) an efficient thermal-driven 3D floorplanning algorithm with an integrated compact resistive network thermal model (CBA-T); (3) two fast thermal-driven 3D floorplanning algorithms using two different thermal models with different runtime and quality (CBA-T-Fast and CBA-T-Hybrid). Our experiments show that the proposed 3D floorplan algorithm with CBA representation can reduce the wirelength by 29% compared with a recent published result from (Hsiu et al., 2004). In addition, compared to a nonthermal-driven 3D floorplanning algorithm, the thermal-driven 3D floorplanning algorithm can reduce the maximum on-chip temperature by 56%.

/pdf/a-thermal-driven-floorplanning-algorithm-for-3d-ics-2e7oixsujx.pdf

A thermal-driven floorplanning algorithm for 3D ICs

Convolutional neural networks (CNNs) have been widely applied in many deep learning applications. In recent years, the FPGA implementation for CNNs has attracted much attention because of its high performance and energy efficiency. However, existing implementations have difficulty to fully leverage the computation power of the latest FPGAs. In this paper we implement CNN on an FPGA using a systolic array architecture, which can achieve high clock frequency under high resource utilization. We provide an analytical model for performance and resource utilization and develop an automatic design space exploration framework, as well as source-to-source code transformation from a C program to a CNN implementation using systolic array. The experimental results show that our framework is able to generate the accelerator for real-life CNN models, achieving up to 461 GFlops for floating point data type and 1.2 Tops for 8-16 bit fixed point.

Jason Cong

Papers

Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks

High-Level Synthesis for FPGAs: From Prototyping to Deployment

FlowMap: an optimal technology mapping algorithm for delay optimization in lookup-table based FPGA designs

A thermal-driven floorplanning algorithm for 3D ICs

Automated Systolic Array Architecture Synthesis for High Throughput CNN Inference on FPGAs