Towards Understanding Sample Variance in Visually Grounded Language Generation: Evaluations and Observations

Open in new window