Novel Image Caption System Using Deep Convolutional Neural Networks (VGG16)
Alaa Sabeeh Salim, Mohammed Basil Abdulkareem, Yarub Essam Fadhel, Abdulkarem Basil Abdulkarem, Ahmed Muhi Shantaf, Ahmed B. Abdulkareem · 2022 International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA) · 2022
With advances in artificial intelligence field and computer vision, image captioning (IC) tool has progressively attracted researchers' attention. IC automatically generates natural text descriptions according to the image content. IC combines the knowledge of computer vision and natural language processing. In this article, a Novel Image Captioning system was developed. The final system was validated on Flicker8K dataset. The novel designed system consists of Long Short Time Memory (LSTM) and VGG16 with Convolution Neural Network (CNN). The main improvements of this system are in structure of designed system by adapting batch size. Also, studying deep learning parameters such as Regularization terms that can be added to loss function, CNN Optimizers and Dropout layer. The results showed the effectiveness of the designed system. Finally, this article highlighted some open challenges in the describing images task.