Konvolüsyonel Sinir Ağına Sahip Akıllı Kontrol Cihazları

Abstract

Bir konvolüsyon sinir ağı (CNN) bir veya daha fazla konvolüsyon katmanları (genellikle bir alt örnekleme) ile oluşur ve ardından bir veya daha fazla tam katman, çok katmanlı sinir ağı olarak izler. CNN mimarisi 2D giriş görüntü (veya konuşma sinyali gibi diğer 2D giriş) yapısının yararlanmak için tasarlanmıştır. Bu sonuç yerel bağlantılar ile elde edilir ve statik parçaların çevirisi toplama sonuçları türü tarafından takip ağırlık ile ilgilidir. CNN'lerin diğer bir avantajı, eğitilmesi daha kolaydır ve aynı sayıda gizli birimle daha az tam ağ parametrelerine sahiptir. Bu tezin amacı, yakalama cihazı (kamera) kullanılarak el hareketlerini otomatik olarak algılayabilen. Konvolüsyonel sinir ağı kullanarak gerçek zamanlı olarak görüntü işleme sistemini tasarlayıp yürütmektir. Program, konvülsiyonel sinir ağıyla elle hareket tespiti yapacak; işitme engelli ve görme engelli insanlara yardım ederek evlerinde cihazlara kontrol etmelerine yardımcı olacak, aynı anda birçok farklı jesti algılayabilecek ve taleplerini yerine getirmek için onlara yardımcı olacaktır. Bu program ile el hareketleri kullanarak elektrik, Mekanik veya elektronik olup olmadığı kontrol edinilebilecektir. Program aynı zamanda işitme engelli insanlara işaret dilini öğretmek için kullanılabilecektir. Keras makine öğrenme Kütüphanesi kullanılarak, CNN modelini eğitmek ve oluşturmak için gerekli yazılım Python’da geliştirilecektir. Ayrıca Python, yakalama cihazından (kamera) görüntü tanıma taleplerini bekleyen istemci tarafındaki soketi çalıştırıyor ve fonksiyonları internet üzerinden çıkış birimine (Arduino) gönderiyordu. Ayrıca 3. parti ortamı (firebase) var, donanımı yazılımla birleştirmek için ortam olarak çalışıyor.

Keywords

• Ardunio,Sign Language,Convolutional Neural Network,Python

Smart Controlling Devices with Convolutional Neural Network

Abstract

A neural network convolutional (CNN) is composed of one or more convolutional layers (often with a step of sub-sampling) and then followed by one or more full layers, as in a neural network multilayer. The architecture of CNN is designed to take advantage of the structure to a 2D input image (or other 2D input such as a speech signal). This result is obtained with local connections and are relevant to the weight, followed by the type of aggregation results in the translation of the static parts. Another advantage of the CNNs so that it is easier to train and they have a fewer number of parameters of the full network with the same number of hidden units. See all the tutorials on convolution and collecting for more details of this operation. The goal of this work is to design and execute a system using Convolutional Neural Network that capable of automatically detect and recognize hand gesture through capture and process image in real time by using capture device (camera) . The program going to be on hand gesture detection with convolutional neural network , and it can be detect many different types of gestures at same time that will be used for helping disabled , deaf or blind people that will helping them to control things in their House and helping them for fulfilling their demands.With this program we can control electrical devices, mechanicals or electronics by using hands gestures . The program also can be used as teaching the deaf peoples to learn sign language. The code for training and creating the CNN model will be in Python using the Keras machine learning.

Keywords

• Ardunio,Sign Language,Convolutional Neural Network

References

1. Anonymous6, 2017, Keras: The Python Deep Learning library, https://keras.io/, [Accessed:12 Des 2018].
2. Anonymous8, 2017, Deep Learning, http://deeplearning.net/software/theano/introduction.htm, [Accessed:15 Jun 2018].
3. Anonymous12, 2016, About CNN, kernels and scale/rotation invariance, https://stats.stackexchange.com/questions/239076/about-cnn-kernels-and-scale-ro, [Accessed:14 Dec 2018].
4. Anonymous13, firebase Realtime Database, https://firebase.google.com/docs/database/, [Accessed:14 Dec 2018].
5. Anonymous14, 2013, Changing Color-space, https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_colorspaces/py_colorspaces.html, [Accessed:31 Dec 2018].
6. Anonymous15, 2013, 2D Convolution ( Image Filtering ), https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_filtering/py_filtering.html, [Accessed:12 Dec 2018].
7. Anonymous16, 2017, Image Thresholding, https://docs.opencv.org/3.4.0/d7/d4d/tutorial_py_thresholding.html, [Accessed:31 Dec 2017].
8. Bouchrika, T., Jemai, O., Zaied, M. ve Amar, C. B., 2014, A new hand posture recognizer based on hybrid wavelet network including a fuzzy decision support system, International Conference on Intelligent Data Engineering and Automated Learning, 183-190.

9. Das, S., 2017, CNN Architectures: LeNet, AlexNet, VGG, GoogLeNet, ResNet and more …. [Online], https://medium.com/@sidereal/cnns-architectures-lenet-alexnet-vgg-googlenet-resnet-and-more-666091488df5, [Accessed:12 Dec 2018].
10. Li, H., Lin, Z., Shen, X., Brandt, J. ve Hua, G., 2015, A convolutional neural network cascade for face detection, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 5325-5334.
11. Liang, M. ve Hu, X., 2015, Recurrent convolutional neural network for object recognition, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 3367-3375.
12. Molchanov, P., Gupta, S., Kim, K. ve Kautz, J., 2015, Hand gesture recognition with 3D convolutional neural networks, Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 1-7.
13. Oliveira, M., Chatbri, H., Little, S., Ferstl, Y., O'Connor, N. E. ve Sutherland, A., 2017, Irish sign language recognition using principal component analysis and convolutional neural networks, Digital Image Computing: Techniques and Applications (DICTA), 2017 International Conference on, 1-8.
14. Rashed, J. R. ve Hasan, H. A., 2017, NEW METHOD FOR HAND GESTURE RECOGNITION USING WAVELET NEURAL NETWORK, Journal of Engineering and Sustainable Development, 21 (1), 65-73.
15. Shi, B., Bai, X. ve Yao, C., 2017, An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition, IEEE Transactions on Pattern Analysis and Machine Intelligence, 39 (11), 2298-2304.
16. Stockman, G. ve Shapiro, L. G., 2001, Computer Vision. 1st, Upper Saddle River, NJ, USA: Prentice Hall PTR.
17. TARRATACA, L., 2008, A gesture recognition System using smartphones, Dissertação (Mestrado em Sistemas de Informação e Engenharia da Computação ….Yadav, P., A deeper understanding of NNets (Part 1) — CNNs [online], https://towardsdatascience.com/a-deeper-understanding-of-nnets-part-1-cnns-263a6e3ac61?fbclid=IwAR1-GhBSZYgGFYfDPN9F02T9eZwowUgRRDnA6p8KOypO73ID68op5de8h0Q, [Accessed:12 Dec 2018].