Object Detection Algorithm Based on Improved Feature Extraction Network

Ting Qiao; Hansong Su; Gaohua Liu; Meng Wang

doi:10.3788/LOP56.231008

Journals >Laser & Optoelectronics Progress >Volume 56 >Issue 23 >Page 231008 > Article

Laser & Optoelectronics Progress
Vol. 56, Issue 23, 231008 (2019)

Object Detection Algorithm Based on Improved Feature Extraction Network

Ting Qiao, Hansong Su, Gaohua Liu^*, and Meng Wang

Author Affiliations

School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China

show less

DOI: 10.3788/LOP56.231008 Cite this Article Set citation alerts

Ting Qiao, Hansong Su, Gaohua Liu, Meng Wang. Object Detection Algorithm Based on Improved Feature Extraction Network[J]. Laser & Optoelectronics Progress, 2019, 56(23): 231008 Copy Citation Text

show less

Fig. 1. Comparison of effects of four data augmentation techniques

Download full size

Fig. 2. Comparison of combination effects of flipping, cropping, and rotating methods

Download full size

Fig. 3. Node representations of cell structures of ResNet, DensNet, and two-path networks. (a) ResNet network; (b) DensNet network; (c) two-path network

Download full size

Fig. 4. Examples of traditional NMS problems. (a) Horses; (b) birds

Download full size

Fig. 5. Trend of parameter quantity of feature extraction network with Top-1 error rate and 52, 100, and 133 layers

Download full size

Fig. 6. Trend of parameter quantity of feature extraction network with Top-1 error rate and network growth rates of 12, 18, 24, and 48

Download full size

Layer	Output size	Detail
Conv1	112×112	7×7,64,stride 2
Conv2	56×56	3×3 max pool,stride 2 $\begin{matrix} [\begin{matrix} 1 \times 1 conv \\ 3 \times 3 conv \\ 1 \times 1 conv \end{matrix}] \end{matrix}$ ×α₁
Conv3	28×28	$\begin{matrix} [\begin{matrix} 1 \times 1 conv \\ 3 \times 3 conv \\ 1 \times 1 conv \end{matrix}] \end{matrix}$ ×α₂
Conv4	14×14	$\begin{matrix} [\begin{matrix} 1 \times 1 conv \\ 3 \times 3 conv \\ 1 \times 1 conv \end{matrix}] \end{matrix}$ ×α₃
Conv5	7×7	$\begin{matrix} [\begin{matrix} 1 \times 1 conv \\ 3 \times 3 conv \\ 1 \times 1 conv \end{matrix}] \end{matrix}$ ×α₄

Table 1. Structure of feature extraction network

Feature extraction network	Depth	Parameter /10⁶
VGG-16	16	168
DensNet(K=48)	161	111
ResNet	101	150
Ours(α₁α₂α₃α₄=6,8,16,3; K=48)	100	134

Table 2. Comparison of complexity of different feature extraction networks

Different parameter	A $\begin{matrix} {P^{0 .}}^{5} \end{matrix}$	A $\begin{matrix} {P^{0 .}}^{5 w} \end{matrix}$	A $\begin{matrix} {P^{0 .}}^{6} \end{matrix}$	A $\begin{matrix} {P^{0 .}}^{6 w} \end{matrix}$	A $\begin{matrix} {P^{0 .}}^{7} \end{matrix}$	A $\begin{matrix} {P^{0 .}}^{7 w} \end{matrix}$
Normal NMS	44.37	44.83	39.18	39.67	29.83	30.34
β=2.5, σ=0.4	46.42	46.92	42.83	43.40	34.68	35.24
β=1.67, σ=0.6	46.58	47.11	43.30	43.79	35.21	35.76
β=1.25, σ=0.8	45.93	46.45	41.68	42.21	33.01	33.53

Table 3. Influences of IoU threshold, β parameter, and weighted average on AP (0.5, 0.6, and 0.7 represent different IoU thresholds; w represents weighted average)

Detection framwork	Backbone	Training set	Testing set	mAP /%
OursNo augmentation No improved NMS	ProposedProposedProposed	VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012	VOC2007VOC2007VOC2007	79.176.678.0

Table 4. Influences of data augmentation and improved NMS mechanism on accuracy

Nums of epoch	Learning rate setting	mAP /%
0	No warming up	78.20
2	0.01, 0.1	78.25
345	0.001, 0.01, 0.10.0001, 0.001, 0.01, 0.10.00001, 0.0001, 0.001, 0.01, 0.1	78.3678.6778.71

Table 5. Influences of different epochs on accuracy

Method		Backbone	Training set	Testing set	mAP/%	Frame rate /(frame·s^-1)
Twostage	Fast R-CNNFaster R-CNNFaster R-CNNMR-CNNIONOurs	VGG-16VGG-16ResNet-101ResNet-101VGG-16Proposed	VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012	VOC2007VOC2007VOC2007VOC2007VOC2007VOC2007	70.073.276.478.276.579.1	0.5072.400.031.252.10
Onestage	YOLOYOLOv2SSD321SSD300*DSOD300DSSD513	GoogleNetDarknet-19ResNet-101VGG-16DS/64-192-48-1ResNet-101	VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012VOC2007+VOC2012	VOC2007VOC2007VOC2007VOC2007VOC2007VOC2007	63.478.677.177.277.781.5	454011.204617.405.50

Table 6. Testing results of different algorithms under VOC2007+VOC2012 training sets

Ting Qiao, Hansong Su, Gaohua Liu, Meng Wang. Object Detection Algorithm Based on Improved Feature Extraction Network[J]. Laser & Optoelectronics Progress, 2019, 56(23): 231008

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information