Inquiry about the similarity results

[dangelosue]

Hey there. :)

I'm curious about the similarity results from associate.py
Based on your code, the motorbike has a similarity of 1.0 with a bicycle but 0 with the rest of the classes.
But even the sofa has a similarity of 0.409 with a bottle...

motorbike [('bicycle', 1.0), ('tvmonitor', 0.0), ('train', 0.0), ('pottedplant', 0.0), ('person', 0.0)]
sofa [('chair', 0.867), ('diningtable', 0.596), ('car', 0.431), ('bottle', 0.409), ('aeroplane', 0.407)]

Do you think it is approvable?

And I also wonder why you guys have done the experiments on all different seeds selectively.

seeds = {
        1: dict(shot1=1, shot2=6, shot3=1, shot5=4, shot10=4),
        2: dict(shot1=5, shot2=5, shot3=1, shot5=3, shot10=0),
        3: dict(shot1=1, shot2=1, shot3=1, shot5=5, shot10=1),
    }