Simple Visual Question Answering

This repository contains Neural network and Demo deploying.

To practice building End-to-end AI application.

Data set: easy-VQA

ConvNet Architecture:

  conv		maxpool			conv 		maxpool		  flatten	  fc

64x64 ---> 64x64x8 ---> 32x32x8 ---> 32x32x16 ---> 16x16x16 ---> 4096 ---> 32

Requirements:

• python 3.7
• torch 1.7
• numpy
• easy_vqa

TODO:

• Train model
• Build demo