luckysunda/Knowledge_aware_VQA_Network

Course project of NLU course

Project E1 246: Natural Language Understanding

• The code contains Pytorch implementation of knowledge-based visual question answering system.

Architecture

Modules

• Fact module
• Question module
• Memory module
• Linear Layer module
• Answer module

Prerequistes

• Pytorch
• numpy
• pickle
• codecs
• sklearn
• easy to execute the entire code in google colab

Dataset

• Preprocessed KG facts (http://malllabiisc.github.io/resources/kvqa/)

• Dataset (http://dosa.cds.iisc.ac.in/kvqa/dataset.json)

Procedure

• It is advised to use google colab.