HiFlipVX: An Open Source High-Level Synthesis FPGA Library for Image Processing

Kalms, Lester; Podlubne, Ariel; Göhringer, Diana

doi:10.1007/978-3-030-17227-5_12

Cited by 19 publications

(21 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The advantage of this approach is that less multiplications and weights are required for the convolution process. Comparing it to a classic 2D convolution, it has a similar effect to the separable filter shown in [17].…”

Section: Depthwise Convolutionmentioning

confidence: 95%

“…High performance with an efficient use of resources can be achieved through the streaming approach and the various parallelization parameters. The integration of the proposed CNN library into the HiFlipVX image processing library [17], which has been extended for object detection [15], increases the range of possible applications that can be implemented in the field of computer vision. Following the OpenVX standard [5] makes it easier to create a heterogeneous system consisting of different architectures (e.g.…”

Section: Related Workmentioning

confidence: 99%

“…HiFlipVX is an open source [16] HLS FPGA library for image processing [17], which has been extended for object detection [15]. The library contains 46 C++ streamingcapable functions, which are mostly based on the OpenVX standard.…”

Section: The Hiflipvx Librarymentioning

confidence: 99%

“…The library functions are parametrizable using C++ templates and highly optimized for performance and FPGA resources. In comparison to the xfOpenCV library from Xilinx [36], it only consumed in average 39% FFs and 32% Lookup Table (LUTs) for a selected set of functions [17]. In addition to the OpenVX standard, most functions support different vectorization options (2x, 4x, 8x) and additional data types (8-, 16-or 32-bit signed/unsigned integers).…”

Section: The Hiflipvx Librarymentioning

confidence: 99%

“…Furthermore, we have researched and implemented different possibilities of parallelization in order to achieve a high performance with an efficient use of resources, which we show in this paper using our implementation of the MobileNets algorithm [9]. Our library is integrated into the HiFlipVX library, which is an open source HLS FPGA library for image processing [17] and object detection [15]. This offers the possibility to design and implement different kinds of computer vision applications with one library.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Parametrizable High-Level Synthesis Library for Accelerating Neural Networks on FPGAs

Kalms

Rad

Ali

et al. 2021

J Sign Process Syst

Self Cite

View full text Add to dashboard Cite

In recent years, Convolutional Neural Network CNN have been incorporated in a large number of applications, including multimedia retrieval and image classification. However, CNN based algorithms are computationally and resource intensive and therefore difficult to be used in embedded systems. FPGA based accelerators are becoming more and more popular in research and industry due to their flexibility and energy efficiency. However, the available resources and the size of the on-chip memory can limit the performance of the FPGA accelerator for CNN. This work proposes an High-Level Synthesis HLS library for CNN algorithms. It contains seven different streaming-capable CNN (plus two conversion) functions for creating large neural networks with deep pipelines. The different functions have many parameter settings (e.g. for resolution, feature maps, data types, kernel size, parallelilization, accuracy, etc.), which also enable compile-time optimizations. Our functions are integrated into the HiFlipVX library, which is an open source HLS FPGA library for image processing and object detection. This offers the possibility to implement different types of computer vision applications with one library. Due to the various configuration and parallelization possibilities of the library functions, it is possible to implement a high-performance, scalable and resource-efficient system, as our evaluation of the MobileNets algorithm shows.

show abstract

Section: Depthwise Convolutionmentioning

confidence: 95%

Section: Related Workmentioning

confidence: 99%