A Lightweight Binarized Convolutional Neural Network Model for Small Memory and Low-Cost Mobile Devices

<div>CBCNN architecture. (a) The size of neural network input is 32 × 32 × 1 on GTSRB. (b) The size of neural network input is 28 × 28 × 1 on fashion-MNIST and MNIST.</div>

Mobile Information Systems

fig2

Figure 2

Figure 2: A Lightweight Binarized Convolutional Neural Network Model for Small Memory and Low-Cost Mobile Devices 