Convolutional Neural Network Based Vehicle Classification in Adverse Illuminous Conditions for Intelligent Transportation Systems

Table 1

Architecture of modified CNN.

Layer name	Output size	Layers
Conv 1	112 × 112	Kernel size = 7 × 7, number of kernels = 64, stride = 2
Pooling	56 × 56	Kernel size = 3 × 3, stride = 2
Conv 2	56 × 56	× 3
Conv 3	28 × 28	× 3
Conv 4	14 × 14	× 3
Conv 5	7 × 7	× 3
Pooling	1 × 1	Adaptive-average-pooling 2d
Proposed classification block		fc1: In-features = 2048, out-feature = 1024
		Relu (in-place)
		drop-out(0.5)
		fc2: In-features = 1024, out-features = 6
		Softmax()
		Classification output (cross entropy)