SPICE
SPICE copied to clipboard
About COMPOSITE dataset in the paper
Hi: I am confused about the COMPOSITE dataset proposed in the paper. How 19985 annotated sentences choose from the source file download from [here] (https://www.dropbox.com/s/z1l4o3j0gj8yyp8/AMT_eval.zip?dl=0). The number of sentences for an image is 4 in 30k and MScoco instead of 3. Thanks~