what does score matching mean in test.py ?
winnechan opened this issue · comments
Winne2GO commented
mil_prob = test_img(im, mil_model, base_image_size, mean)
mil_prob = mil_prob.data.cpu().float().numpy()
...
for jj in xrange(prec.shape[1]):
f = interp1d(precision_score['thresh'][jj][:,0], precision_score['prec'][jj][:,0])
#prec[:, jj] = f(mil_prob[:, jj])
prec[:, jj] = mil_prob[:, jj]
mil_prec = prec
the code seems only copy values in mil_prob to mil_prec ? the value shown in the example, what does that mean ? why need the coco_valid1_eval.pkl ? what does the content of coco_valid1_eval.pkl mean ?
i am confused, can someone help ?