Novel Shoe Type Recognition Method Based on Convolutional Neural Network

Fig. 6. Photographs of partial misidentify of shoe type. (a) Example of shoe image with label 4 being incorrectly identified as label 14; (b) example of shoe image with label 3 being incorrectly identified as label 19

Download full size

Table 1. Effect of number of output elements in Ip1 layer on performance
View table
Table 1. Effect of number of output elements in Ip1 layer on performance
Number of output elements 300 500 1000
Test accuracy /% 89.68 91.91 93.46
Train time /min 48 50 56

Table 2. Effect of network depth on performance
View table
Table 2. Effect of network depth on performance
k_{ernel_size} Number of layers Memory /MB Train time /min Accuracy /% Loss
5×5 6 83.6 50 91.91 0.2807
3×3 8 150.4 78 95.81 0.1601

Table 3. Effect of overlapping pooling on performance
View table
Table 3. Effect of overlapping pooling on performance
Pooling Memory /MB Traintime /min Accuracy /%
Original pooling 150.4 78 95.81
Overlapping pooling 145.9 72 96.06

Tools

Get Citation

Copy Citation Text

Mengjing Yang, Yunqi Tang, Xiaojia Jiang. Novel Shoe Type Recognition Method Based on Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(19): 191505

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites