卷积神经网络在物体检测方面的应用

The applicationof convolutional neural network in object detection

深度学习的应用越来越广泛,而且在图像处理方面有着很好的效果。而其中的卷积神经网络(convolutional neural network,CNN)就是一个很好的应用典型,虽然卷积神经网络在特征提取方面有很大的优势,并且也取得了不错的效果。但是在某些具体的检测任务,卷积神经网络(convolutional neural network,CNN)会面临如何处理在图像提取特征的时候,如何保证特征不畸变的问题。本文将通过对卷积神经网络模型的改进来进一步分析如何处理这个问题。实验结果表明,我们提出的多特征池化层能够很好地改善CNN的这一不足。

The application of deep learning is more and more extensive and has a good effect on image processing？ While the convolutional neural network（convolutional neural network, CNN） is a typical application,although the convolution neural network has a great advantage in feature extraction, it also has a good effect？ But in some specific testing tasks, convolution neural network（convolutional neural network, CNN） will face how to deal with at the time of image feature extraction, how to ensure the characteristic distortion problem？ This paper will further analyze how to deal with this problem by improving the convolution neural network model？ The experimental results show that the multi-feature pooling layer can improve CNN’s deficiency？