Extracting Image Features Through Deep Learning
Abstract
The purpose of this study is to identify images with deep learning with the least error. In machine learning projects, the basis of the work is extracting features from raw data. Finally, we differentiate different features through classifiers. In the present project, images with dimensions of 224*224 are applied to the network. Most networks use color images, which have 3 channels, the final dimensions of which are 3*224*224. We used the vgg19 network to extract the feature from the image with the highest accuracy. To increase the speed of weight correction operations, batch_size = 30 is considered. 70% of the images were used for network training, 20% for validation and 10% of the data for network testing and evaluation. The speed and accuracy of this project is high.
Ā
References
[2] Courville, I. G. a. Y. B. a. A., Deep Learning: MIT Press, 2016.
[3] Bengio, Y., Courville, A., & Vincent, P., 2013, "Representation learning: A review and new perspectives". IEEE transactions on pattern analysis and machine intelligence, 35(8), 1798-1828.
[4] Bengio, Y., 2009, "Learning deep architectures for AI". Foundations and trendsĀ® in Machine Learning, 2(1), 1-127.
[5] Hinton, G. E., & Salakhutdinov, R. R., 2006, "Reducing the dimensionality of data with neural networks". Science, 313(5786), 504-507.
[6] Lee, H., Grosse, R., Ranganath, R., & Ng, A. Y., 2009, "Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations", Proceedings of the 26th annual international conference on machine learning.
[7] Liu, J. N., Hu, Y., You, J. J., & Chan, P. W., 2014, "Deep neural network based feature representation for weather forecasting", Proceedings on the International Conference on Artificial Intelligence (ICAI).
[8] Mehdi Jamaseb Khalari,V. Derhami, "Identify hand mode in video with deep learning", Sixth Joint Congress of Fuzzy Systems and Ho Shamand Iran March 2010.
[9] http://www.7khatcode.com.