Exploring Structural Sparsity in CNN via Selective Penalty

Zhao, Mingxin; Peng, Junbo; Yu, Shuangming; Liu, Liyuan; Wu, Nanjian

doi:10.1109/tcsvt.2021.3071532

Cited by 18 publications

(4 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Later on, several techniques for compressing models were proposed. These techniques discard certain connections or features within the networks to reduce unnecessary parameters and computational expenses [20][21][22][23].…”

Section: Model Compression Techniquesmentioning

confidence: 99%

“…The model pruning approach [20] removes the less important parameters according to their impact on feature extraction performance. The model structural sparsity approach [21] learns sparse parameter matrices to reduce computational complexity during model optimisation. The weight quantisation approach [22] uses fewer bits to represent the network weights to reduce memory consumption.…”

Section: Model Compression Techniquesmentioning

confidence: 99%

“…Existing strategies for lightweight deep learning models consist of model pruning [20], structural sparsity [21], quantisation [22], and distillation [23]. For HSI classification, designing lightweight CNNs while maintaining strong feature extraction capabilities has attracted the interest of researchers.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

LiteCCLKNet: A lightweight criss‐cross large kernel convolutional neural network for hyperspectral image classification

Zhong

Gong²,

Zhang

et al. 2023

IET Computer Vision

View full text Add to dashboard Cite

High‐performance convolutional neural networks (CNNs) stack many convolutional layers to obtain powerful feature extraction capability, which leads to huge storing and computational costs. The authors focus on lightweight models for hyperspectral image (HSI) classification, so a novel lightweight criss‐cross large kernel convolutional neural network (LiteCCLKNet) is proposed. Specifically, a lightweight module containing two 1D convolutions with self‐attention mechanisms in orthogonal directions is presented. By setting large kernels within the 1D convolutional layers, the proposed module can efficiently aggregate long‐range contextual features. In addition, the authors effectively obtain a global receptive field by stacking only two of the proposed modules. Compared with traditional lightweight CNNs, LiteCCLKNet reduces the number of parameters for easy deployment to resource‐limited platforms. Experimental results on three HSI datasets demonstrate that the proposed LiteCCLKNet outperforms the previous lightweight CNNs and has higher storage efficiency.

show abstract

Section: Model Compression Techniquesmentioning

confidence: 99%

Section: Model Compression Techniquesmentioning

confidence: 99%

See 1 more Smart Citation

LiteCCLKNet: A lightweight criss‐cross large kernel convolutional neural network for hyperspectral image classification

Zhong

Gong²,

Zhang

et al. 2023

IET Computer Vision

View full text Add to dashboard Cite

show abstract

“…In order to improve the accuracy and quantity of vehicle detection, networks with higher accuracy are needed. However, due to the huge amount of network parameters and computation, these networks will bring high resource and computation costs [ 15 ]. How to reduce the parameter amount and algorithm complexity to improve the acceleration performance becomes one of the difficulties.…”

Section: Introductionmentioning

confidence: 99%

FPGA-Based Vehicle Detection and Tracking Accelerator

Zhai

et al. 2023

Sensors

View full text Add to dashboard Cite

A convolutional neural network-based multiobject detection and tracking algorithm can be applied to vehicle detection and traffic flow statistics, thus enabling smart transportation. Aiming at the problems of the high computational complexity of multiobject detection and tracking algorithms, a large number of model parameters, and difficulty in achieving high throughput with a low power consumption in edge devices, we design and implement a low-power, low-latency, high-precision, and configurable vehicle detector based on a field programmable gate array (FPGA) with YOLOv3 (You-Only-Look-Once-version3), YOLOv3-tiny CNNs (Convolutional Neural Networks), and the Deepsort algorithm. First, we use a dynamic threshold structured pruning method based on a scaling factor to significantly compress the detection model size on the premise that the accuracy does not decrease. Second, a dynamic 16-bit fixed-point quantization algorithm is used to quantify the network parameters to reduce the memory occupation of the network model. Furthermore, we generate a reidentification (RE-ID) dataset from the UA-DETRAC dataset and train the appearance feature extraction network on the Deepsort algorithm to improve the vehicles’ tracking performance. Finally, we implement hardware optimization techniques such as memory interlayer multiplexing, parameter rearrangement, ping-pong buffering, multichannel transfer, pipelining, Im2col+GEMM, and Winograd algorithms to improve resource utilization and computational efficiency. The experimental results demonstrate that the compressed YOLOv3 and YOLOv3-tiny network models decrease in size by 85.7% and 98.2%, respectively. The dual-module parallel acceleration meets the demand of the 6-way parallel video stream vehicle detection with the peak throughput at 168.72 fps.

show abstract