Recent Advances in Convolutional Neural Networks for Object Recognition

DOI

Bibliographic Information

Other Title
  • 物体認識のための畳み込みニューラルネットワークの研究動向

Description

Since the advent of AlexNet in the image recognition competition ILSVRC'12, convolution neural networks (CNN) has become the de facto standard approach in image recognition. In ILSVRC, new CNN models have been proposed every year, and have consistently contributed to the improvement of image recognition accuracy. CNN has been widely used not only in image classification but also as a base network for various deep learning tasks such as segmentation and object detection. In this paper, we overview the evolution of CNN after AlexNet, survey about various advanced CNN models proposed in recent years, and classify them into several approaches and explain. In addition, comprehensive evaluation are performed using publicly available datasets for representative models, and tendencies of accuracy and training time of each model are discussed.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top