Pretraining Convolutional Neural Networks for Image-Based Vehicle Classification

Table 2

Accuracy and FPS of different methods.

Method	Accuracy	FPS
HOG+SVM	60.12%	4
DAISY+SVM	69.04%	2
ORB+BoW+SVM	64.07%	7
SIFT+BoW+SVM	74.49%	5
DeCAF[1]	66.20%	13
CNNs	91.42%	800