Infrared object detection network compression using Lp normalized weight

被引：0
|
作者：
Li W. [1 ]
Yang X. [1 ]
Li C. [1 ]
Lu R. [1 ]
Xie X. [1 ]
He C. [1 ]
机构：
[1] Institute of Missile Engineering, Rocket Force Engineering University, Xi'an
来源：
Hongwai yu Jiguang Gongcheng/Infrared and Laser Engineering | 2021年 / 50卷 / 08期
关键词：
Constrained gradient descent; Infrared object detection; Lp normalization; Sparse neural network;
D O I：
10.3788/IRLA20200510
中图分类号：
学科分类号：
摘要：
In view of the characteristic that the infrared image has less texture compared with RGB image, an infrared object detection network compression method using Lp normalized weight was proposed. It aimed at improving the adaptability of convolutional neural network based object detection framework to the infrared images, and compressing the scale of network while improving its generalization ability. Firstly, the phenomenon that the sparsity of Lp normalized weight can be precisely controlled by adjusting p was revealed. Based on the phenomenon, a sparsification method for object detection network was proposed. It respectively trained the backbone network and the detector with Lp spherical gradient descent and classical gradient descent, to balance the network scale and fitting accuracy. The tests on simulated infrared image dataset show that, the proposed method is superior to the dense model on both of network scale and detection accuracy: in terms of network scale, the sparsification reduces the effective parameters of Faster R-CNN, Single Shot multibox Detector (SSD) and YOLOv3 by 52%, 78% and 66% respectively; it also improves the mean Average Precision (mAP) of Faster R-CNN, SSD and YOLOv3 by 0.1%, 0.3% and 0.2%, thus verifying the effectiveness of the proposed method. © 2021, Editorial Board of Journal of Infrared and Laser Engineering. All right reserved.
引用
收藏
相关论文
共 14 条
  • [1] Lienhart R, Maydt J., An extended set of Haar-like features for rapid object detection, International Conference on Image Processing, (2002)
  • [2] Dalal N, Triggs B., Histograms of oriented gradients for human detection, IEEE Computer Society Conference on Computer Vision & Pattern Recognition, (2005)
  • [3] Lowe D G., Distinctive image features from scale-invariant keypoints, International Journal of Computer Vision, 60, 2, pp. 91-110, (2004)
  • [4] Bay H, Ess A, Tuytelaars T, Et al., Speeded-up robust features, Computer Vision and Image Understanding, 110, 3, pp. 346-359, (2008)
  • [5] Li X, Wang L, Sung E., AdaBoost with SVM-based component classifiers, Engineering Applications of Artificial Intelligence, 21, 5, pp. 785-795, (2008)
  • [6] Felzenszwalb P F, Huttenlocher D P., Pictorial structures for object recognition, International Journal of Computer Vision, 61, 1, pp. 55-79, (2005)
  • [7] Felzenszwalb P F, Girshick R B, McAllester D., Cascade object detection with deformable part models, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 2241-2248, (2010)
  • [8] Felzenszwalb P F, Girshick R B, Mcallester D A., Visual object detection with deformable part models, The Twenty-Third IEEE Conference on Computer Vision and Pattern Recognition, (2010)
  • [9] Zhang Xiuling, Hou Daibiao, Zhang Chengcheng, Et al., Design of MPCANet fire image recognition model for deep learning, Infrared and Laser Engineering, 47, 2, (2018)
  • [10] Gong Junliang, He Xin, Wei Zhonghui, Et al., Infrared dim and small target detection method using scale-space theory, Infrared and Laser Engineering, 42, 9, pp. 2566-2573, (2013)
12