Coverage for src/bob/bio/face/database/vgg2.py: 91%
35 statements
« prev ^ index » next coverage.py v7.6.0, created at 2024-07-13 00:04 +0200
1#!/usr/bin/env python
2# vim: set fileencoding=utf-8 :
3# Tiago de Freitas Pereira <tiago.pereira@idiap.ch>
5import logging
7logger = logging.getLogger(__name__)
10"""
11 VGG2 database implementation
12"""
14from clapper.rc import UserDefaults
15from sklearn.base import BaseEstimator, TransformerMixin
16from sklearn.pipeline import make_pipeline
18import bob.io.base
20from bob.bio.base.database import CSVDatabase, FileSampleLoader
21from bob.pipelines import DelayedSample
23rc = UserDefaults("bobrc.toml")
26class VGG2Annotations(TransformerMixin, BaseEstimator):
27 """
28 Decode VGG2 Annotations.
30 VGG2 has 5 landmarks annotated and the are the following: two for eyes, one for the nose and two for the mouth
32 """
34 def fit(self, X, y=None):
35 return self
37 def _more_tags(self):
38 return {
39 "requires_fit": False,
40 }
42 def transform(self, X):
43 """
44 Convert leye_x, leye_y, reye_x, reye_y attributes to `annotations = (leye, reye)`
45 """
47 annotated_samples = []
49 for x in X:
50 annotations = {
51 "leye": (
52 float(x.leye_y),
53 float(x.leye_x),
54 ),
55 "reye": (
56 float(x.reye_y),
57 float(x.reye_x),
58 ),
59 "nose": (
60 float(x.nose_y),
61 float(x.nose_x),
62 ),
63 "lmouth": (
64 float(x.lmouth_y),
65 float(x.lmouth_x),
66 ),
67 "rmouth": (
68 float(x.rmouth_y),
69 float(x.rmouth_x),
70 ),
71 "topleft": (
72 float(x.face_y),
73 float(x.face_x),
74 ),
75 "size": (
76 float(x.face_h),
77 float(x.face_w),
78 ),
79 }
81 sample = DelayedSample.from_sample(x, annotations=annotations)
82 # Cleaning up
83 [
84 delattr(sample, a)
85 for a in [
86 "leye_x",
87 "leye_y",
88 "reye_x",
89 "reye_y",
90 "nose_y",
91 "nose_x",
92 "face_y",
93 "face_x",
94 "face_h",
95 "face_w",
96 "lmouth_y",
97 "lmouth_x",
98 "rmouth_y",
99 "rmouth_x",
100 ]
101 ]
102 annotated_samples.append(sample)
104 return annotated_samples
107class VGG2Database(CSVDatabase):
108 """
109 The VGG2 Dataset is composed of 9131 people split into two sets.
110 The training set contains 8631 identities, while the test set contains 500 identities.
112 As metadata, this dataset contains the gender labels "m" and "f" for, respectively, male and female.
113 It also contains the following race labels:
115 - A: Asian in general (Chinese, Japanese, Filipino, Korean, Polynesian, Indonesian, Samoan, or any other Pacific Islander
116 - B: A person having origins in any of the black racial groups of Africa
117 - I: American Indian, Asian Indian, Eskimo, or Alaskan native
118 - U: Of indeterminable race
119 - W: Caucasian, Mexican, Puerto Rican, Cuban, Central or South American, or other Spanish culture or origin, Regardless of race
120 - N: None of the above
122 Race labels are taken from: MasterEBTSv10.0.809302017_Final.pdf.
124 This dataset also contains sets for `T-Norm` and `Z-Norm`, normalization.
127 We provide four protocols; `vgg2-short`, `vgg2-full`,`vgg2-short-with-eval`, `vgg2-full-with-eval`.
128 The `vgg2-short` and `vgg2-full` present the sample amount of identities but
129 varies with respect to the number of samples per identity.
130 The `vgg2-full` preserves the number of samples per identity from the original dataset.
131 On the other hand, the `vgg2-short` presents 10 samples per identity at the probe and training sets.
132 With that the training set of `vgg2-short` contains 86'310 samples instead of 3'141'890 samples
133 from `vgg2-full`.
134 The protocols with the suffix `-with-eval`, splits the original test set into a dev and eval sets
135 containing 250 identities each.
138 All the landmarks and face crops provided in the original dataset is provided with this interface.
140 .. warning::
142 To use this dataset protocol, you need to have the original files of the VGG2 dataset.
143 Once you have it downloaded, please run the following command to set the path for Bob
145 .. code-block:: sh
147 bob config set bob.bio.face.vgg2.directory [VGG2 PATH]
148 bob config set bob.bio.face.vgg2.extension [VGG2 EXTENSION]
150 For more information check:
152 .. code-block:: latex
154 @inproceedings{cao2018vggface2,
155 title={Vggface2: A dataset for recognising faces across pose and age},
156 author={Cao, Qiong and Shen, Li and Xie, Weidi and Parkhi, Omkar M and Zisserman, Andrew},
157 booktitle={2018 13th IEEE international conference on automatic face \\& gesture recognition (FG 2018)},
158 pages={67--74},
159 year={2018},
160 organization={IEEE}
161 }
162 """
164 name = "vgg2"
165 category = "face"
166 dataset_protocols_name = "vgg2.tar.gz"
167 dataset_protocols_urls = [
168 "https://www.idiap.ch/software/bob/databases/latest/face/vgg2-8c067663.tar.gz",
169 "http://www.idiap.ch/software/bob/databases/latest/face/vgg2-8c067663.tar.gz",
170 ]
171 dataset_protocols_hash = "8c067663"
173 def __init__(
174 self,
175 protocol,
176 dataset_original_directory=rc.get("bob.bio.face.vgg2.directory", ""),
177 dataset_original_extension=rc.get(
178 "bob.bio.face.vgg2.extension", ".jpg"
179 ),
180 annotation_type="eyes-center",
181 fixed_positions=None,
182 ):
183 super().__init__(
184 name=self.name,
185 protocol=protocol,
186 transformer=make_pipeline(
187 FileSampleLoader(
188 data_loader=bob.io.base.load,
189 dataset_original_directory=dataset_original_directory,
190 extension=dataset_original_extension,
191 ),
192 VGG2Annotations(),
193 ),
194 templates_metadata=["gender", "race"],
195 annotation_type=annotation_type,
196 fixed_positions=fixed_positions,
197 )
199 def background_model_samples(self):
200 if self.protocol == "vgg2-full":
201 logger.warning(
202 "This set is very long (3M samples). It might take ~4 minutes to load everything"
203 )
204 return super().background_model_samples()