The dataset contains 112,085 screen summaries in CSV format for 22,417 screens from the RICO dataset. The split/ folder contains the train/dev/test splits used by the paper for model training and evaluation.
google-research-datasets/screen2words
The dataset includes screen summaries that describes Android app screenshot's functionalities. It is used for training and evaluation of the screen2words models (our paper accepted by UIST'21 will be linked soon).