Source code for quaterion.dataset.similarity_samples
from dataclasses import dataclass
from typing import Any, Union
[docs]@dataclass
class SimilarityPairSample:
"""Represents a pair of objects, their similarity and relationship with other pairs.
Examples::
data = [
# First query group (subgroup)
SimilarityPairSample(
obj_a="cheesecake",
obj_b="muffins",
score=0.9,
subgroup=10
),
SimilarityPairSample(
obj_a="cheesecake",
obj_b="macaroons",
score=0.8,
subgroup=10
),
SimilarityPairSample(
obj_a="cheesecake",
obj_b="candies",
score=0.7,
subgroup=10
),
# Second query group (subgroup)
SimilarityPairSample(
obj_a="lemon",
obj_b="lime",
score=0.9,
subgroup=11
),
SimilarityPairSample(
obj_a="lemon",
obj_b="orange",
score=0.7,
subgroup=11
),
]
"""
obj_a: Any
obj_b: Any
score: float = 1.0
"""Similarity score between Object A and B.
It is assumed, that `score = 1.0` - means objects are similar,
`score = 0.0` objects are completely different.
"""
subgroup: int = 0
"""Consider all examples outside this group as negative samples.
By default, all samples belong to group 0 - therefore other samples could
not be used as negative examples."""
[docs]@dataclass
class SimilarityGroupSample:
"""Represent groups of similar objects all of which should match with one-another
within the group.
Examples::
Faces dataset.
All pictures of a single person should have single unique group
id. In this case NN will learn to match all pictures within
the group closer to each-other, but pictures from different
groups - further.
file_name group_id
0 elon_musk_1.jpg 555
1 elon_musk_2.jpg 555
2 elon_musk_3.jpg 555
3 leonard_nimoy_1.jpg 209
4 leonard_nimoy_2.jpg 209
"""
obj: Any
group: int
SimilaritySample = Union[SimilarityGroupSample, SimilarityPairSample]
