Image scene understanding by using Recurrent Neural Networks
| dc.contributor.author | Aldabergen А. | |
| dc.date.accessioned | 2024-12-20T10:08:27Z | |
| dc.date.available | 2024-12-20T10:08:27Z | |
| dc.date.issued | 2020 | |
| dc.description.abstract | This work proposes an implementation of Recurrent Neural Networks (RNN) for image scene understanding. Task is clear: given an image and the system should provide an accurate description for the given image. The novelty of the work is that this system is realized on Telegram Bot. Fine tuned model learns where to look, its focus is shifted across the image by the help of attention mechanism. Thus the model was able to find the most relevant parts of the image and find out most relevant words that describe the scene. It has an encoder-decoder architecture. As own contribution, transfer learning is implemented on pre-trained model. The significance of the work is that this kind of system can be easily implemented in bunch of areas of our life rather than other capturing applications. | |
| dc.identifier.uri | https://repository.sdu.edu.kz/handle/123456789/1602 | |
| dc.language.iso | en | |
| dc.subject | Recurrent Neural Networks,Telegram Bot, mechanism, architecture, model, applications | |
| dc.title | Image scene understanding by using Recurrent Neural Networks | |
| dc.type | Other |