الفهرس | Only 14 pages are availabe for public view |
Abstract A set of deep learning models and various datasets were tested to solve the problem by finding the link between the image features and the words it represented. The work was divided into two phases: the first was to extract features and determine classes. Various models were tested on different datasets (ImageNet, MS-COCO) to determine the effectiveness of their use. Combination of ALEXNET network, multi-class SVM was the best with accuracy 84.25%.The second was to generate captions, by entering the features and classes from the first stage. Various models were tested, and concluded that LSTM was the best model.The two phases resulted in a hybrid model of ALEXNET network, multi-class SVM and LSTM as the best model with accuracy 88.4%.The model was tested on the complete MS-COCO dataset, reaching an accuracy 90.7%, and was shown to reduce image processing time and high accuracy compared to previous models |