Visual Question Answering System
| dc.contributor.author | Phukan, Doli | |
| dc.date.accessioned | 2026-10-08T05:23:55Z | |
| dc.date.available | 2026-10-08T05:23:55Z | |
| dc.date.issued | 2022 | |
| dc.description.abstract | We can easily observe an image and answer any inquiry about it using our common sense understanding as humans. However, there are also other instances, such as a visually challenged person or an intelligence analyst attempting to extract visual information from an image is very challenging. Hence, an AI system called Visual Question Answering (VQA) is developed. VQA is an artificial intelligence system that takes an image and a query about it as input and output an answer. Computer Vision (CV), Natural Language Processing (NLP), and Knowledge Representation (KR) are the basic building blocks of VQA. There are various open issues in VQA. One of them is sometimes a model can predict ‘Yes’ or ‘No’ as an answer which is not relatable for the question which requires a descriptive answer and vice versa. To solve this issue in VQA domain, we incorporate Question Segregation (QS) technique to classify the questions into three types (‘Yes/No’, ‘Other’ and ’Number’). Then we successfully in corporated this technique with two of the VQA models, SAN and MCAN. We evaluate the performance of the QS and SAN model on two datasets VQA v.2 and CLEVR. We also studied and analysed the impact of Question Segregation on the performance of all these two models on different datasets | |
| dc.identifier.uri | http://nits.ndl.gov.in/handle/123456789/176 | |
| dc.language.iso | en | |
| dc.publisher | Department of Computer Science & Engineering, NITS | |
| dc.relation.ispartofseries | D1543 | |
| dc.title | Visual Question Answering System | |
| dc.type | Thesis |