A Lightweight Model for Traffic Sign Classification Based on Enhanced LeNet-5 Network

Table 7

Accuracy of our proposed network compared with that of previous state-of-the-art approaches and 2 CNN models tested by us (model: CNN_1, model: classic_LeNet_5) on Belgian Traffic Sign Data Set (BTSD).

Methods	Accuracy (%)
Single CNN with 3 STNs [47]	98.87
One CNN [48]	98.17
INNLP+SRC (PI) [49]	97.83
Small CNN [46]	98.1
Model: CNN_1	90.43
Model: classic_LeNet_5	92.81
Model: modified_LeNet_5 (our proposed network)	98.37