This supplementary material includes four folders: code and data
Below is a brief description of each folder.

1. code
This folder contains the scripts used for the experiments presented in the paper.
GQA-Q2Q.zip: Source code for all experiments.
Note: The code requires images and scene graphs from the original GQA dataset.

2. data
This folder contains the initial GQA-Q2Q dataset and machine-annotated samples.
GQA_Q2Q_train.csv: Training set of the initial GQA-Q2Q dataset
GQA_Q2Q_valid.csv: Validation set of the initial GQA-Q2Q dataset
GQA_Q2Q_test.csv: Test set of the initial GQA-Q2Q dataset
GQA_Q2Q_machine_annotation_sample.csv: 150 randomly sampled machine-annotated examples (limited subset of the full annotations)
undefined_Q2Q_sample.csv: 1,000 ambiguous question samples for machine annotation (limited set due to size constraints)