Reference no: EM132371770
Assignment -
1. System contains two parts knowledge generation from images using YOLO (images to be taken from GQA dataset)
2. Converting English query into prolog facts into java (queries to be taken GQA dataset) (note that there are multiple reasoning or inference queries, java system should be able to convert such queries into prolog)
Part 1 - Using YOLO generate knowledge facts (prolog facts) from the images and feed it into prolog system.
Part 2 - A java system that converts English question from GQA dataset into prolog facts. Read paper AAAI_20 which gives a reference. I am expecting similar java code as I have sent in javaFiles.zip. These java files corresponds to CASPR.pdf where the user answers query from textual passages. We have to answer queries from visual images.
a. At first start with a dummy knowledge base and convert English queries into prolog facts using java code.
b. Fit it into a prolog system. Prolog system should find the answer for the query given the dummy knowledge file.
c. I have attached dhruvTest.txt to show how such files look like. These files are fed into SASP engine. SASP is an open source software build on prolog. You can use prolog for this purpose if you don't know syntax of SASP.
d. Follow the architecture in figure 2 in AAAI_20 paper. We will follow the same architecture but with a different dataset e.g. GQA.