دانلود رایگان مقاله انگلیسی طبقه بندی برش های معده براساس یادگیری عمیق و نمایش تنک - IEEE 2018

عنوان فارسی
طبقه بندی برش های معده براساس یادگیری عمیق و نمایش تنک
عنوان انگلیسی
Classification of Gastric Slices based on Deep Learning and Sparse Representation
صفحات مقاله فارسی
0
صفحات مقاله انگلیسی
5
سال انتشار
2018
نشریه
آی تریپل ای - IEEE
فرمت مقاله انگلیسی
PDF
کد محصول
E8670
رشته های مرتبط با این مقاله
مهندسی کامپیوتر
گرایش های مرتبط با این مقاله
هوش مصنوعی
مجله
کنفرانس کنترل و تصمیم گیری در چین - Chinese Control And Decision Conference
دانشگاه
Harbin Institute of Technology - School of Computer and Technology - Harbin
کلمات کلیدی
شبکه عصبی پیچشی، تجزیه تنک، ماشین بردار پشتیبانی، تصویر شکاف معده
چکیده

Abstract:


In this paper, a classification method based on deep learning and sparse representation is proposed for gastric slice images. First of all, the convolution features of images are extracted through the convolutional neural network. Then an overcomplete learning dictionary of convolution features can be gained by K-SVD algorithm. Convolution features can obtain their sparse representation through decomposition on the overcomplete dictionary. Finally, a linear SVM classifier is used to classify the sparse representation of convolution features instead of the commonly used nonlinear classifiers, so as to achieve the purpose of classifying gastric slices. Experiments show that using linear SVM reduces the computation cost of classifier and achieves better classification effect than convolutional neural network with sigmoid output layer or SVM with RBF kernel. Therefore, the method proposed in this paper has better classification effect and lower computation cost.

نتیجه گیری

 CONCLUSION


In this paper, an image classification method based on deep learning and sparse representation is proposed for the classification of gastric slice images. The convolution features of the images are extracted by CNN. An overcomplete learning dictionary of these features can be then learned through K-SVD algorithm. The convolution features are decomposed on overcomplete dictionary to get their sparse representation. Finally, the best classification result can be obtained by only using linear SVM, not nonlinear. This method reduces the computational complexity of the classifier while obtaining the best classification effect, which may make the classification method work in real time.


بدون دیدگاه