Image Classification Based on TensorFlow and Convolution Neural Networks

Zhuobao Tang ； Hanyun Wen

《International Journal of Social Science and Education Research》 5卷3期 (2022/03) Pp. 28-34

https://doi.org/10.6918/IJOSSER.202203_5(3).0007

摘要

It is no longer a fresh thing to use convolutional neural networks to classify images. The most common experiment is to use CNN for handwritten digit recognition. However, the pictures used for training in the recognition just have two colors of black and white while which mostly are colored in the real world. As a result, more complex models need to be designed for training and classification. This article mainly does the following work: Study the principle of CNN and its application on graph classification; Use TensorFlow to quickly build a classification model; Apply the classification model to cifar10 for graph classification and evaluated the classification model. The accuracy of the classifier designed is 78% on the training set and 70% on the verification set.

關鍵字

TensorFlow ； Convolutional neural network ； Graph classification

參考文獻

Weston J, Ratle F, Mobahi H: Neural networks: Tricks of the trade (Springer, Berlin, 2012), p.639-655.

C. Szegedy et al.: Going deeper with convolutions, Boston, 2015 IEEE Conference on Computer Vision and Pattern Recognition (Boston, MA, 2015), Vol. 1, p.1-9.

Fukunaga K, Narendra P M: A branch and bound algorithm for computing k-nearest neighbors. IEEE Transactions on Computers, Vol. 100 (1975) No.7, p.750-753.

Breiman L: Random forests. Machine Learning, Vol. 45 (2001) No.1, p. 5-32.

Vincent P, Larochelle H, Lajoie I: Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion, Journal of Machine Learning Research Vol. 11 (2010) No.12, p.3371-3408.