def predicts(predictor, start, end):
global hit, total_hit
lines = f[start:end]
imgs = [line.split('\t')[0] for line in lines]
img_features = np.array([[float(x) for x in line.split('\t')[1:]]
for line in lines])
score = predictor.bulk_predict(img_features, evaluator.all_distinct_texts)
for i, img in enumerate(imgs):
indexes = (-score[i]).argsort()
if FLAGS.print_predict:
evaluator.print_img(img, i)
evaluator.print_neareast_texts_with_indexes(score[i], indexes, 20)
evaluator.print_neareast_texts(rerank_score[i])
has_hit = False
for i in range(FLAGS.topn):
if evaluator.all_distinct_text_strs[indexes[i]] in m[img]:
total_hit += 1
has_hit = True
if has_hit:
hit += 1
print('hits:', hit)
print('total_hits:', total_hit)
def predicts(predictor, start, end):
lines = f[start:end]
imgs = [line.split('\t')[0] for line in lines]
img_features = np.array([[float(x) for x in line.split('\t')[1:]]
for line in lines])
score = bulk_predict(predictor, img_features, all_distinct_texts)
#print(score)
for i, img in enumerate(imgs):
indexes = (-score[i]).argsort()
if FLAGS.print_predict:
#print(i, img)
evaluator.print_img(img, i)
evaluator.print_neareast_texts_from_sorted(score[i], indexes[:10],
img)
#evaluator.print_neareast_texts(score[i], num=20, img = None)
hits = img2text[img]
labels = [indexes[j] in hits for j in xrange(num_texts)]
rank_metrics.add(labels)