【摘要】 卷积神经网络在图像处理领域取得了突出的表现,但是由于其庞大的计算量使得它的应用范围受到限制。通常,卷积层的计算量占据了整个网络的大部分计算,主要包含有大量的乘法和加法,本文针对卷积层的计算特点,实现了一种高效的卷积层加速模块的设计。最后通过实验结果表明,在计算相同的网络结构下,该设计相比于CPU的计算效率更高。更多还原

【Abstract】 The convolutional neural network has achieved great success in image processing field, but its application is limited because of its huge computational burden. Usually, the computational part of convolutional layer occupies most of the computation in the network, mainly including a large number of multiplications and additions. An efficient design of convolution layer, which makes full use of this feature, is proposed in this paper. Finally, the experimental results show that the design is more efficient than CPU under the same network structure.更多还原