Story Generation from Images using Deep Learning
Students names: Abrar Alname
Supervised by Dr. Miada Almasre and Dr. nora Almalki
-
-
-
-
Recently, the problem of creating descriptive captions for images became a significant one. However, the expressivity of human languages which highlights the (intent, emotion, and idea) had been among the challenges which hindered researchers from widely experimenting with creating linguistically rich captions for images. That motivated us to utilize advanced deep learning algorithms to generate expressive (stylistically rich) captions for images.
The main objective of this research is to develop an AI prototype which can generate short literary texts in English based on sentence-length, image captions; basically, an intelligent system capable of captioning images (describing them as humans would do) and using them to create smart literary content.
- Develop a Neural Image Processing which utilizes an encoder-decoder architecture to generate sentence-level captions of images.
- Develop a Neural Text Generator which utilizes the encoder-decoder architecture to generate literary (expressive) story in English based on the output of the image captioning decoder.
- Evaluate the performance of both encoder-decoder models using the BLEU-4 metrics, and human evaluators.
A prototype of an AI system that can generate short literary stories in English based on image captions through using deep learning algorithms and natural language processing techniques. We will train both an image processing and a text generator as the two main components of the system.
The proposed model will be evaluated on the Flickr8k dataset using the python platform, enhanced with installations of TensorFlow and Keras libraries. The performance of the proposed model is evaluated using BLEU matric, which outperforms previous benchmark models.
|
Last Update
6/4/2023 11:36:03 AM
|
|
|
|