Modulating Bottom-Up and Top-Down Visual Processing via Language-Conditional Filters
2020·Arxiv