Qualitative Results:

We compare our framework against two SOTA architectures, MVT and SAT, across different amounts of training data; 10%, 40%, 70%, and 100% as shown in the figure above