SUBMISSIONS

SUBMISSION DETAIL

Mehmet Bahadr ETNKAYA, Eda GALP
 


Keywords:



IMAGE RECOGNITION BASED ON DEEP LEARNING
 
Computer-aided image recognition has become an important are of research in recent years with the improvements in the field of artificial intelligence. Deep learning is one of the most robust approaches that is able to produce accurate results in processing of high-dimensional images. In this work, the deep learning based analysis have been carried out on the TensorFlow open source machine learning library. An image data set is produced from the ImageNet database and then deep learning based analysis were performed for the network models of VGG16, VGG19, ResNet50, ResNet5022, ResNet101, MobileNet, DenseNet121 and InceptionResNetV2. When the results obtained have been compared in terms of the accuracy metric, it is seen that the VGG19 produces better accuracy results to that of other network models. The value of the accuracy metric for VGG19 is obtained as 0,9815 which corresponds to %98,15 accuracy. However, the accuracy value for the InceptionResNetV2 network model has been obtained as 0,1273 which corresponds to the worst result among all algorithms. The accuracy rate of %12,73 proves that the InceptionResNetV2 network model can not effectively be used in ImageNet database. On the other hand, the CPU time parameter values for the VGG19 and InceptionResNetV2 have respectively been obtained as 6,830 and 2,976 seconds. As a result, it can be expressed that VGG19 network model can successfully be used in recognition of images taken from the ImageNet database while the InceptionResNetV2 network model produces worse performance for the same database. ORCID NO: 0000-0003-3378-4561

Anahtar Kelimeler: Image recognition, Deep learning, Network models