Algorithm for Student Behavior Detection Based on Neural Network

Layer	Output size	Ksize	Padding	Stride
Conv 1	100×100	5×5	Same	1
Conv 2	50×50	2×2	Valid	2
Conv 3	50×50	5×5	Same	1
Max pool	25×25	2×2	Valid	2
Conv 4	25×25	3×3	Same	1
Conv 5	12×12	3×3	Valid	2
Conv 6	12×12	3×3	Same	1
Conv 7	6×6	2×2	Valid	2
Fc 1	1152
Fc 2	192
Fc 3	5

Table 2. Accuracies of different network structures under different activation functions and loss functions unit: %
View table
Table 2. Accuracies of different network structures under different activation functions and loss functions unit: %
Network structure Prelu h-Swish bh-Swish
SEUCLID Relative steep loss SEUCLID Relative steep loss SEUCLID Relative steep loss
Original MTCNN 75.3 75.9 77.6 77.8 78.2 78.3
Improved MTCNN 75.8 76.2 77.8 78.1 78.4 78.6

Table 3. Accuracy of each algorithm unit: %
View table
Table 3. Accuracy of each algorithm unit: %
Algorithm YOLOv3 tiny-YOLOv3 SSD MobileNetV2-SSD MobileNetV3-SSD Our algorithm
Dataset 78.2 66.7 79.2 75.8 77.2 78.6
Jetson TX2 74.5 60.4 77.1 73.3 74.1 77.9

Table 4. Average time of each algorithm to process 1 frame image on Jetson TX2 unit: ms
View table
Table 4. Average time of each algorithm to process 1 frame image on Jetson TX2 unit: ms
Algorithm YOLOv3 tiny-YOLOv3 SSD MobileNetV2-SSD MobileNetV3-SSD Our algorithm
Time 780 88 653 80 76 69

Tools

Get Citation

Copy Citation Text

Hansong Su, Tengteng Liu, Gaohua Liu, Xichu Tian. Algorithm for Student Behavior Detection Based on Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221016

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Mar. 19, 2020

Accepted: Apr. 20, 2020

Published Online: Oct. 30, 2020

The Author Email: Gaohua Liu (suppig@126.com)

DOI:10.3788/LOP57.221016

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Network structure for image classification

Table 1. Network structure for image classification

Table 2. Accuracies of different network structures under different activation functions and loss functions unit: %

Table 2. Accuracies of different network structures under different activation functions and loss functions unit: %

Table 3. Accuracy of each algorithm unit: %

Table 3. Accuracy of each algorithm unit: %

Table 4. Average time of each algorithm to process 1 frame image on Jetson TX2 unit: ms

Table 4. Average time of each algorithm to process 1 frame image on Jetson TX2 unit: ms