Bindctnet: A Simple Binary Dct Network for Image Classification.

Xiangrui Xing,Xian Yu,Wenao Ma,Yue Huang,Delu Zeng,Xinghao Ding
DOI: https://doi.org/10.1109/icassp.2018.8461941
2018-01-01
Abstract:Convolution neural networks play an important role in the image classification tasks. However, it is time consuming to train the network and the cost of memory resources is usually high. In this paper, a simple and effective network named BinDCTNet is presented by using the binary discrete cosine transform(BinDCT) to extract the feature-maps and a hyper-parameter to reduce dimension of the extracted feature. The proposed network has extremely low computing complexity and there is almost no parameters needed to be stored. Experiments are carried out on the hand written digit dataset MNIST and the vehicle logo VLOGO dataset. The results show that the proposed network achieves the state-of-the-art accuracy with fast speed and low memory cost, which makes it applicable on mobile and embedded devices.
What problem does this paper attempt to address?