Video Question Answering Systems(VQA)
- 0 Collaborators
A lot of research has been done on Text-based Question Answering Systems. However, very little work was done on answering questions whose answer lies in the image until 2016. This was mainly, due to the locality of the answer in image lies in a certain part of the image which needs to be focused. After 2016 various attention mechanism based QA were proposed to solve this problem. However, the accuracy still remains low for open-ended questions. Moreover, a complex model needs to be designed for a system that intends to answer questions based on a video. Here, sequence-to-sequence modeling is required for both, CNN features as well as text-based features. Integrating these features is a major complex challenge which I intend to tackle. ...learn more
Project status: Under Development
Intel Technologies
AI DevCloud / Xeon,
Intel Opt ML/DL Framework
Overview / Usage
Under Development
Methodology / Approach
Tentative: Use sequence-to-sequence modeling to get features from video and text.
Technologies Used
python3, keras, pytorch. (Intel version of these softwares and packages optimized for DevCloud)