diff --git a/tests/conftest.py b/tests/conftest.py
index 50f1348c60268206f780a8188d9754429ebd73e9..53bf4aae33aae318169e0539df533a80e369c01f 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -101,9 +101,7 @@ def mock_database(tmp_path_factory):
element_json = json.loads(element_path.read_text())
element_type = element_json["type"]
- image_path = (
- FIXTURES / "extraction" / "images" / element_type / id
- ).with_suffix(".jpg")
+ image_path = (FIXTURES / "extraction" / "images" / id).with_suffix(".jpg")
polygon = element_json.get("polygon")
# Always use page images because polygons are based on the full image
diff --git a/tests/data/entities.yml b/tests/data/entities.yml
index 4bf7513a40cbeec0b6ce5367fcf0925848277d79..a10f027eaa31efa33796d0f678d096c09e003a41 100644
--- a/tests/data/entities.yml
+++ b/tests/data/entities.yml
@@ -4,7 +4,7 @@ entities:
- Arkindex's entity
- Person /!\
- adj
-- birthdate
+- age
- firstname
- fullname
- name
diff --git a/tests/data/extraction/elements/test-page_1-line_1.json b/tests/data/extraction/elements/test-page_1-line_1.json
index 716bf50aa37442665d7ebc7c9246ac3b0bf01739..de3fb8f02e86a6263d6835eb47047bf98130935f 100644
--- a/tests/data/extraction/elements/test-page_1-line_1.json
+++ b/tests/data/extraction/elements/test-page_1-line_1.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [37, 191],
- [37, 339],
- [767, 339],
- [767, 191],
- [37, 191]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Coupez",
+ "name": "Leunaut",
"type": "surname",
"offset": 0
},
{
- "name": "Bouis",
+ "name": "Claude",
"type": "firstname",
- "offset": 7
+ "offset": 8
},
{
- "name": "7.12.14",
- "type": "birthdate",
- "offset": 13
+ "name": "49",
+ "type": "age",
+ "offset": 15
}
]
}
diff --git a/tests/data/extraction/elements/test-page_1-line_2.json b/tests/data/extraction/elements/test-page_1-line_2.json
index c3d79c187b6879135757cea304b1fb78cdf9be49..bbc58881db63b48d2d00abaef877cd4ae200bc24 100644
--- a/tests/data/extraction/elements/test-page_1-line_2.json
+++ b/tests/data/extraction/elements/test-page_1-line_2.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [28, 339],
- [28, 464],
- [767, 464],
- [767, 339],
- [28, 339]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Poutrain",
+ "name": "Bauracho",
"type": "surname",
"offset": 0
},
{
- "name": "Adolphe",
+ "name": "Claudine",
"type": "firstname",
"offset": 9
},
{
- "name": "9.4.13",
- "type": "birthdate",
- "offset": 17
+ "name": "39",
+ "type": "age",
+ "offset": 18
}
]
}
diff --git a/tests/data/extraction/elements/test-page_1-line_3.json b/tests/data/extraction/elements/test-page_1-line_3.json
index 78fbe7865e47e062dc7b2b9682edc337b3294e53..61a71912e665fecc77ce032623654e63b91c888f 100644
--- a/tests/data/extraction/elements/test-page_1-line_3.json
+++ b/tests/data/extraction/elements/test-page_1-line_3.json
@@ -1,26 +1,26 @@
{
"type": "text_line",
"polygon": [
- [28, 464],
- [28, 614],
- [767, 614],
- [767, 464],
- [28, 464]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Gabale",
+ "name": "Laurent",
"type": "surname",
"offset": 0
},
{
- "name": "Français",
+ "name": "Jacquse",
"type": "firstname",
- "offset": 7
+ "offset": 8
},
{
- "name": "26.3.11",
- "type": "birthdate",
+ "name": "21",
+ "type": "age",
"offset": 16
}
]
diff --git a/tests/data/extraction/elements/test-page_2-line_1.json b/tests/data/extraction/elements/test-page_2-line_1.json
index 9d5c131b90c8b1bd52cd6e365899c8f43afb070e..dfa3554e4e5037b6846dcc01d5d3e88299f52b86 100644
--- a/tests/data/extraction/elements/test-page_2-line_1.json
+++ b/tests/data/extraction/elements/test-page_2-line_1.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [14, 199],
- [14, 330],
- [767, 330],
- [767, 199],
- [14, 199]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Durosoy",
+ "name": "Valette",
"type": "surname",
"offset": 0
},
{
- "name": "Bouis",
+ "name": "Elisabeth",
"type": "firstname",
"offset": 8
},
{
- "name": "22-4-18",
- "type": "birthdate",
- "offset": 14
+ "name": "76",
+ "type": "age",
+ "offset": 18
}
]
}
diff --git a/tests/data/extraction/elements/test-page_2-line_2.json b/tests/data/extraction/elements/test-page_2-line_2.json
index 9eda1ee8b10d4e98be32548b1d3b30caa3c3ce98..af3d79cec4b43a3fb1709c61adff61ae0502c02c 100644
--- a/tests/data/extraction/elements/test-page_2-line_2.json
+++ b/tests/data/extraction/elements/test-page_2-line_2.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [16, 330],
- [16, 471],
- [765, 471],
- [765, 330],
- [16, 330]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Colaiani",
+ "name": "Tanbol",
"type": "surname",
"offset": 0
},
{
- "name": "Angels",
+ "name": "Jean",
"type": "firstname",
- "offset": 9
+ "offset": 7
},
{
- "name": "28.11.17",
- "type": "birthdate",
- "offset": 16
+ "name": "76",
+ "type": "age",
+ "offset": 12
}
]
}
diff --git a/tests/data/extraction/elements/test-page_2-line_3.json b/tests/data/extraction/elements/test-page_2-line_3.json
index e372adcd8a65c853c8f1ced78d9f352003fda4a4..52c5b510eb7e3ab6fc8a258d53de69433b96b061 100644
--- a/tests/data/extraction/elements/test-page_2-line_3.json
+++ b/tests/data/extraction/elements/test-page_2-line_3.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [11, 473],
- [11, 598],
- [772, 598],
- [772, 473],
- [11, 473]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Renouard",
+ "name": "Vauret",
"type": "surname",
"offset": 0
},
{
- "name": "Maurice",
+ "name": "Jean",
"type": "firstname",
- "offset": 9
+ "offset": 7
},
{
- "name": "25.7.04",
- "type": "birthdate",
- "offset": 17
+ "name": "64",
+ "type": "age",
+ "offset": 12
}
]
}
diff --git a/tests/data/extraction/elements/train-page_1-line_1.json b/tests/data/extraction/elements/train-page_1-line_1.json
index 907dd2a040b31fb96997a9aff856cf7d0a77ad1a..4a39286284ab35f1f2691a1eead776468b982269 100644
--- a/tests/data/extraction/elements/train-page_1-line_1.json
+++ b/tests/data/extraction/elements/train-page_1-line_1.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [27, 187],
- [27, 327],
- [754, 327],
- [754, 187],
- [27, 187]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Caillet",
+ "name": "Laulont",
"type": "surname",
"offset": 0
},
{
- "name": "Maurice",
+ "name": "Francois",
"type": "firstname",
"offset": 8
},
{
- "name": "28.9.06",
- "type": "birthdate",
- "offset": 16
+ "name": "8",
+ "type": "age",
+ "offset": 17
}
]
}
diff --git a/tests/data/extraction/elements/train-page_1-line_2.json b/tests/data/extraction/elements/train-page_1-line_2.json
index 080806424555c66cce1af9a38ed064d3332795cf..cdc9ec5a754ccb84d2fc8e5ec6286a4780da5502 100644
--- a/tests/data/extraction/elements/train-page_1-line_2.json
+++ b/tests/data/extraction/elements/train-page_1-line_2.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [28, 328],
- [28, 465],
- [755, 465],
- [755, 328],
- [28, 328]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Reboul",
+ "name": "Ciret",
"type": "surname",
"offset": 0
},
{
- "name": "Jean",
+ "name": "Antoine",
"type": "firstname",
- "offset": 7
+ "offset": 6
},
{
- "name": "30.9.02",
- "type": "birthdate",
- "offset": 12
+ "name": "27",
+ "type": "age",
+ "offset": 14
}
]
}
diff --git a/tests/data/extraction/elements/train-page_1-line_3.json b/tests/data/extraction/elements/train-page_1-line_3.json
index 63b38d499ba311497acd8100454c9d1074cd478f..70387cc1590e49cbaff2032982d749239e4fdb44 100644
--- a/tests/data/extraction/elements/train-page_1-line_3.json
+++ b/tests/data/extraction/elements/train-page_1-line_3.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [23, 463],
- [23, 604],
- [803, 604],
- [803, 463],
- [23, 463]
+ [0, 0],
+ [0, 85],
+ [1900, 85],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Bareyre",
+ "name": "Ciret",
"type": "surname",
"offset": 0
},
{
- "name": "Jean",
+ "name": "Marie",
"type": "firstname",
- "offset": 8
+ "offset": 6
},
{
- "name": "28.3.11",
- "type": "birthdate",
- "offset": 13
+ "name": "28",
+ "type": "age",
+ "offset": 12
}
]
}
diff --git a/tests/data/extraction/elements/train-page_1-line_4.json b/tests/data/extraction/elements/train-page_1-line_4.json
index eb348b19e2bd57869d73a89778fb1d1e9879c221..0efa34f9a517785a473f7d63b28a8154bf793f6b 100644
--- a/tests/data/extraction/elements/train-page_1-line_4.json
+++ b/tests/data/extraction/elements/train-page_1-line_4.json
@@ -1,26 +1,26 @@
{
"type": "text_line",
"polygon": [
- [21, 604],
- [21, 743],
- [812, 743],
- [812, 604],
- [21, 604]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Roussy",
+ "name": "Ciret",
"type": "surname",
"offset": 0
},
{
- "name": "Jean",
+ "name": "Marie",
"type": "firstname",
- "offset": 7
+ "offset": 6
},
{
- "name": "4.11.14",
- "type": "birthdate",
+ "name": "2",
+ "type": "age",
"offset": 12
}
]
diff --git a/tests/data/extraction/elements/train-page_2-line_1.json b/tests/data/extraction/elements/train-page_2-line_1.json
index e7f6663b4faeb70b677f67332d44c2e28d6f605e..586d9492bac257406fb101996547e777e396d900 100644
--- a/tests/data/extraction/elements/train-page_2-line_1.json
+++ b/tests/data/extraction/elements/train-page_2-line_1.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [18, 197],
- [18, 340],
- [751, 340],
- [751, 197],
- [18, 197]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Marin",
+ "name": "Eureston",
"type": "surname",
"offset": 0
},
{
- "name": "Marcel",
+ "name": "Solange",
"type": "firstname",
- "offset": 6
+ "offset": 9
},
{
- "name": "10.8.06",
- "type": "birthdate",
- "offset": 13
+ "name": "10",
+ "type": "age",
+ "offset": 17
}
]
}
diff --git a/tests/data/extraction/elements/train-page_2-line_2.json b/tests/data/extraction/elements/train-page_2-line_2.json
index bc6829dab843ea9b68a2bbdac8df91c575ebc96d..450c1ebc67f95b4d5b855752d59cb95fc82c53a2 100644
--- a/tests/data/extraction/elements/train-page_2-line_2.json
+++ b/tests/data/extraction/elements/train-page_2-line_2.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [18, 340],
- [18, 476],
- [751, 476],
- [751, 340],
- [18, 340]
+ [0, 0],
+ [0, 83],
+ [1900, 83],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Amical",
+ "name": "Terontussieux",
"type": "surname",
"offset": 0
},
{
- "name": "Eloi",
+ "name": "Jean",
"type": "firstname",
- "offset": 7
+ "offset": 14
},
{
- "name": "11.10.04",
- "type": "birthdate",
- "offset": 12
+ "name": "2",
+ "type": "age",
+ "offset": 19
}
]
}
diff --git a/tests/data/extraction/elements/train-page_2-line_3.json b/tests/data/extraction/elements/train-page_2-line_3.json
index 90432163e7449c6c228eb25d1024fc23eec988ca..fd3390582ffda15e92ed1673c469f114f74cf7e3 100644
--- a/tests/data/extraction/elements/train-page_2-line_3.json
+++ b/tests/data/extraction/elements/train-page_2-line_3.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [21, 476],
- [21, 615],
- [746, 615],
- [746, 476],
- [21, 476]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Biros",
+ "name": "Pressonet",
"type": "surname",
"offset": 0
},
{
- "name": "Mael",
+ "name": "Marie",
"type": "firstname",
- "offset": 6
+ "offset": 10
},
{
- "name": "30.10.10",
- "type": "birthdate",
- "offset": 11
+ "name": "12",
+ "type": "age",
+ "offset": 16
}
]
}
diff --git a/tests/data/extraction/elements/val-page_1-line_1.json b/tests/data/extraction/elements/val-page_1-line_1.json
index d255644499a7858d72b4e636f0471b61abd19c93..be1ab12473937cc5f848340d5d8a30465c4c19e5 100644
--- a/tests/data/extraction/elements/val-page_1-line_1.json
+++ b/tests/data/extraction/elements/val-page_1-line_1.json
@@ -1,27 +1,27 @@
{
"type": "text_line",
"polygon": [
- [14, 211],
- [14, 347],
- [755, 347],
- [755, 211],
- [14, 211]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Monard",
+ "name": "Ciraud",
"type": "surname",
"offset": 0
},
{
- "name": "Bouis",
+ "name": "Antoine",
"type": "firstname",
"offset": 7
},
{
- "name": "29-7-04",
- "type": "birthdate",
- "offset": 13
+ "name": "34",
+ "type": "age",
+ "offset": 15
}
]
}
diff --git a/tests/data/extraction/elements/val-page_1-line_2.json b/tests/data/extraction/elements/val-page_1-line_2.json
index 633ccf9b3047dc85bab971adb0dafdc03765d044..a51724d6cff364e7fc4bf158408d6acc641006e0 100644
--- a/tests/data/extraction/elements/val-page_1-line_2.json
+++ b/tests/data/extraction/elements/val-page_1-line_2.json
@@ -1,26 +1,26 @@
{
"type": "text_line",
"polygon": [
- [14, 350],
- [14, 484],
- [748, 484],
- [748, 350],
- [14, 350]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "Astier",
+ "name": "Ciraud",
"type": "surname",
"offset": 0
},
{
- "name": "Arthur",
+ "name": "Priser",
"type": "firstname",
"offset": 7
},
{
- "name": "11-2-13",
- "type": "birthdate",
+ "name": "34",
+ "type": "age",
"offset": 14
}
]
diff --git a/tests/data/extraction/elements/val-page_1-line_3.json b/tests/data/extraction/elements/val-page_1-line_3.json
index 951431c82753e689558577571a84fcf396cf903c..fd868e76abd506b1f0b2ce4cb19198a90e21bad1 100644
--- a/tests/data/extraction/elements/val-page_1-line_3.json
+++ b/tests/data/extraction/elements/val-page_1-line_3.json
@@ -1,26 +1,26 @@
{
"type": "text_line",
"polygon": [
- [11, 484],
- [11, 622],
- [751, 622],
- [751, 484],
- [11, 484]
+ [0, 0],
+ [0, 84],
+ [1900, 84],
+ [1900, 0],
+ [0, 0]
],
"transcription_entities": [
{
- "name": "De Vlieger",
+ "name": "Ciraud",
"type": "surname",
"offset": 0
},
{
- "name": "Jules",
+ "name": "Elisabeth",
"type": "firstname",
- "offset": 11
+ "offset": 7
},
{
- "name": "21-11-11",
- "type": "birthdate",
+ "name": "34",
+ "type": "age",
"offset": 17
}
]
diff --git a/tests/data/extraction/images/double_page/test-page_1.jpg b/tests/data/extraction/images/double_page/test-page_1.jpg
deleted file mode 100644
index bb873b5cf0dba6d7cf4d81b388799e6079d5d0ac..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/double_page/test-page_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/double_page/test-page_2.jpg b/tests/data/extraction/images/double_page/test-page_2.jpg
deleted file mode 100644
index d5282347582f881a5bdc1071638afe9403e26689..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/double_page/test-page_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/double_page/train-page_1.jpg b/tests/data/extraction/images/double_page/train-page_1.jpg
deleted file mode 100644
index ee1a8d546f9633930b51e105bc77bfd7d0524fd6..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/double_page/train-page_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/double_page/train-page_2.jpg b/tests/data/extraction/images/double_page/train-page_2.jpg
deleted file mode 100644
index 6f9b6499f240e00c67e6204c9e6453dc70d684fe..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/double_page/train-page_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/double_page/val-page_1.jpg b/tests/data/extraction/images/double_page/val-page_1.jpg
deleted file mode 100644
index 07a1912cd67099f64e052cbc1a4e89ebb3a83cce..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/double_page/val-page_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/test-page_1-line_1.jpg b/tests/data/extraction/images/test-page_1-line_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a2c1e672403f61c530782d644dc6af1c1de19e0a
Binary files /dev/null and b/tests/data/extraction/images/test-page_1-line_1.jpg differ
diff --git a/tests/data/extraction/images/test-page_1-line_2.jpg b/tests/data/extraction/images/test-page_1-line_2.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9df6239780ac6cde37c3c7327f7183ea481eb02b
Binary files /dev/null and b/tests/data/extraction/images/test-page_1-line_2.jpg differ
diff --git a/tests/data/extraction/images/test-page_1-line_3.jpg b/tests/data/extraction/images/test-page_1-line_3.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b31308f68422bd2a94007ce215f7deddf95d80be
Binary files /dev/null and b/tests/data/extraction/images/test-page_1-line_3.jpg differ
diff --git a/tests/data/extraction/images/test-page_2-line_1.jpg b/tests/data/extraction/images/test-page_2-line_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8f26a3a2def5ad45b3f7e94a0093ac26b84204aa
Binary files /dev/null and b/tests/data/extraction/images/test-page_2-line_1.jpg differ
diff --git a/tests/data/extraction/images/test-page_2-line_2.jpg b/tests/data/extraction/images/test-page_2-line_2.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8639a8f4fcc67b8b986ed8877141a96decf5f169
Binary files /dev/null and b/tests/data/extraction/images/test-page_2-line_2.jpg differ
diff --git a/tests/data/extraction/images/test-page_2-line_3.jpg b/tests/data/extraction/images/test-page_2-line_3.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..abe1e6fcedc735e6248503deddb23a93f79642de
Binary files /dev/null and b/tests/data/extraction/images/test-page_2-line_3.jpg differ
diff --git a/tests/data/extraction/images/text_line/test-page_1-line_1.jpg b/tests/data/extraction/images/text_line/test-page_1-line_1.jpg
deleted file mode 100644
index 59b4173da998b2a6c9a00c6fde7a550920e06b47..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_1-line_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/test-page_1-line_2.jpg b/tests/data/extraction/images/text_line/test-page_1-line_2.jpg
deleted file mode 100644
index aa4c1f0f923882eebca307773d1099864de0622a..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_1-line_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/test-page_1-line_3.jpg b/tests/data/extraction/images/text_line/test-page_1-line_3.jpg
deleted file mode 100644
index bee0d3159564d31de40a21c553d1486cbe296333..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_1-line_3.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/test-page_2-line_1.jpg b/tests/data/extraction/images/text_line/test-page_2-line_1.jpg
deleted file mode 100644
index 44d4f37898a62d2f4bee272cb35add685b06dfdf..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_2-line_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/test-page_2-line_2.jpg b/tests/data/extraction/images/text_line/test-page_2-line_2.jpg
deleted file mode 100644
index c19ec187bbf254e934cd344422ef32731885dba6..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_2-line_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/test-page_2-line_3.jpg b/tests/data/extraction/images/text_line/test-page_2-line_3.jpg
deleted file mode 100644
index 4e05ada4d1c1ce67a85847782cf030245a21b3e2..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/test-page_2-line_3.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_1-line_1.jpg b/tests/data/extraction/images/text_line/train-page_1-line_1.jpg
deleted file mode 100644
index 6768c8183e841c5326facfce7bb4b1bdf5801eff..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_1-line_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_1-line_2.jpg b/tests/data/extraction/images/text_line/train-page_1-line_2.jpg
deleted file mode 100644
index b787d7072945c314c0917c63b3329a78b12bbf97..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_1-line_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_1-line_3.jpg b/tests/data/extraction/images/text_line/train-page_1-line_3.jpg
deleted file mode 100644
index b8ae4811a419bb25d27bd1c684522eeb71e314fa..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_1-line_3.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_1-line_4.jpg b/tests/data/extraction/images/text_line/train-page_1-line_4.jpg
deleted file mode 100644
index 41ee3f51abf4926f2d9c930c7fb3e339171687bc..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_1-line_4.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_2-line_1.jpg b/tests/data/extraction/images/text_line/train-page_2-line_1.jpg
deleted file mode 100644
index a1d15adc504778437929c9e3f244b5b6bcacde77..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_2-line_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_2-line_2.jpg b/tests/data/extraction/images/text_line/train-page_2-line_2.jpg
deleted file mode 100644
index bc7ff4c07f9f6d4765f1d145e6b6e4544a4c31b1..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_2-line_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/train-page_2-line_3.jpg b/tests/data/extraction/images/text_line/train-page_2-line_3.jpg
deleted file mode 100644
index 90e737d16d6cf982cbfe25242985c0da7ee7bc6a..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/train-page_2-line_3.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/val-page_1-line_1.jpg b/tests/data/extraction/images/text_line/val-page_1-line_1.jpg
deleted file mode 100644
index 5937da29f438c8e98494b29a5f549b0b398e3723..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/val-page_1-line_1.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/val-page_1-line_2.jpg b/tests/data/extraction/images/text_line/val-page_1-line_2.jpg
deleted file mode 100644
index 758e2c89b1cd4b37ba2c03a984b417970c384daa..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/val-page_1-line_2.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/text_line/val-page_1-line_3.jpg b/tests/data/extraction/images/text_line/val-page_1-line_3.jpg
deleted file mode 100644
index 30e2e4319acdb5038fc1da831091018ff871c329..0000000000000000000000000000000000000000
Binary files a/tests/data/extraction/images/text_line/val-page_1-line_3.jpg and /dev/null differ
diff --git a/tests/data/extraction/images/train-page_1-line_1.jpg b/tests/data/extraction/images/train-page_1-line_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..579006eeadeda909af6b156b3abcc79be39ad389
Binary files /dev/null and b/tests/data/extraction/images/train-page_1-line_1.jpg differ
diff --git a/tests/data/extraction/images/train-page_1-line_2.jpg b/tests/data/extraction/images/train-page_1-line_2.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c043342cf169f18f4a937b7cb26953111c730c0b
Binary files /dev/null and b/tests/data/extraction/images/train-page_1-line_2.jpg differ
diff --git a/tests/data/extraction/images/train-page_1-line_3.jpg b/tests/data/extraction/images/train-page_1-line_3.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..36140494757daf7717ab616219c30551d146dddc
Binary files /dev/null and b/tests/data/extraction/images/train-page_1-line_3.jpg differ
diff --git a/tests/data/extraction/images/train-page_1-line_4.jpg b/tests/data/extraction/images/train-page_1-line_4.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..42c688826ae22837cd1b3b5d00704721b899e7fc
Binary files /dev/null and b/tests/data/extraction/images/train-page_1-line_4.jpg differ
diff --git a/tests/data/extraction/images/train-page_2-line_1.jpg b/tests/data/extraction/images/train-page_2-line_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1ce1c370792b5df8144ee98a9b66df68a8383069
Binary files /dev/null and b/tests/data/extraction/images/train-page_2-line_1.jpg differ
diff --git a/tests/data/extraction/images/train-page_2-line_2.jpg b/tests/data/extraction/images/train-page_2-line_2.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bfaa7425d60d6b9afd8d56f0a6c115bb92e53edd
Binary files /dev/null and b/tests/data/extraction/images/train-page_2-line_2.jpg differ
diff --git a/tests/data/extraction/images/train-page_2-line_3.jpg b/tests/data/extraction/images/train-page_2-line_3.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..eab65d4b9b02348ff9bfd7b17d4c079835238d8e
Binary files /dev/null and b/tests/data/extraction/images/train-page_2-line_3.jpg differ
diff --git a/tests/data/extraction/images/val-page_1-line_1.jpg b/tests/data/extraction/images/val-page_1-line_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..daac765839dcb6a04269dbeee8cd05fac889dc16
Binary files /dev/null and b/tests/data/extraction/images/val-page_1-line_1.jpg differ
diff --git a/tests/data/extraction/images/val-page_1-line_2.jpg b/tests/data/extraction/images/val-page_1-line_2.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0e2399b4b59449de2596f4aabc731a468ae53c9f
Binary files /dev/null and b/tests/data/extraction/images/val-page_1-line_2.jpg differ
diff --git a/tests/data/extraction/images/val-page_1-line_3.jpg b/tests/data/extraction/images/val-page_1-line_3.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3580fdb558916ddb0be8b2344ac9dc83da807eb1
Binary files /dev/null and b/tests/data/extraction/images/val-page_1-line_3.jpg differ
diff --git a/tests/data/extraction/split.json b/tests/data/extraction/split.json
index e264f689b2e4299e263971869452bb4b7e67544e..143073abfea01d2f351cbc030dbbc905e886b01d 100644
--- a/tests/data/extraction/split.json
+++ b/tests/data/extraction/split.json
@@ -3,470 +3,470 @@
"test-page_1-line_1": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_1-line_1.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_1-line_1.jpg",
"polygon": [
[
- 37,
- 191
+ 0,
+ 0
],
[
- 37,
- 339
+ 0,
+ 83
],
[
- 767,
- 339
+ 1900,
+ 83
],
[
- 767,
- 191
+ 1900,
+ 0
],
[
- 37,
- 191
+ 0,
+ 0
]
]
},
- "text": "â“¢Couâ‡e⇠ⓕBouis â“‘â‡.12.14"
+ "text": "â“¢Leunaut â“•Clauâ‡e â“‘â‡â‡"
},
"test-page_1-line_2": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_1-line_2.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_1-line_2.jpg",
"polygon": [
[
- 28,
- 339
+ 0,
+ 0
],
[
- 28,
- 464
+ 0,
+ 83
],
[
- 767,
- 464
+ 1900,
+ 83
],
[
- 767,
- 339
+ 1900,
+ 0
],
[
- 28,
- 339
+ 0,
+ 0
]
]
},
- "text": "â“¢â‡outrain â“•Aâ‡olâ‡â‡e â“‘9.4.13"
+ "text": "â“¢â‡auracâ‡o â“•Clauâ‡ine â“‘â‡â‡"
},
"test-page_1-line_3": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_1-line_3.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_1-line_3.jpg",
"polygon": [
[
- 28,
- 464
+ 0,
+ 0
],
[
- 28,
- 614
+ 0,
+ 83
],
[
- 767,
- 614
+ 1900,
+ 83
],
[
- 767,
- 464
+ 1900,
+ 0
],
[
- 28,
- 464
+ 0,
+ 0
]
]
},
- "text": "â“¢â‡abale â“•â‡ranâ‡ais â“‘26.3.11"
+ "text": "â“¢Laurent â“•Jacâ‡use â“‘21"
},
"test-page_2-line_1": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_2-line_1.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_2-line_1.jpg",
"polygon": [
[
- 14,
- 199
+ 0,
+ 0
],
[
- 14,
- 330
+ 0,
+ 83
],
[
- 767,
- 330
+ 1900,
+ 83
],
[
- 767,
- 199
+ 1900,
+ 0
],
[
- 14,
- 199
+ 0,
+ 0
]
]
},
- "text": "â“¢â‡urosoy â“•Bouis â“‘22â‡4â‡18"
+ "text": "â“¢â‡alette â“•Elisaâ‡et⇠ⓑ7â‡"
},
"test-page_2-line_2": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_2-line_2.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_2-line_2.jpg",
"polygon": [
[
- 16,
- 330
+ 0,
+ 0
],
[
- 16,
- 471
+ 0,
+ 84
],
[
- 765,
- 471
+ 1900,
+ 84
],
[
- 765,
- 330
+ 1900,
+ 0
],
[
- 16,
- 330
+ 0,
+ 0
]
]
},
- "text": "â“¢Colaiani â“•Anâ‡els â“‘28.11.1â‡"
+ "text": "â“¢Tanâ‡ol â“•Jean â“‘7â‡"
},
"test-page_2-line_3": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/test-page_2-line_3.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/test-page_2-line_3.jpg",
"polygon": [
[
- 11,
- 473
+ 0,
+ 0
],
[
- 11,
- 598
+ 0,
+ 83
],
[
- 772,
- 598
+ 1900,
+ 83
],
[
- 772,
- 473
+ 1900,
+ 0
],
[
- 11,
- 473
+ 0,
+ 0
]
]
},
- "text": "â“¢Renouar⇠ⓕMaurice â“‘2â‡.â‡.04"
+ "text": "â“¢â‡auret â“•Jean â“‘â‡â‡"
}
},
"train": {
"train-page_1-line_1": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_1-line_1.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_1-line_1.jpg",
"polygon": [
[
- 27,
- 187
+ 0,
+ 0
],
[
- 27,
- 327
+ 0,
+ 84
],
[
- 754,
- 327
+ 1900,
+ 84
],
[
- 754,
- 187
+ 1900,
+ 0
],
[
- 27,
- 187
+ 0,
+ 0
]
]
},
- "text": "â“¢Caillet â“•Maurice â“‘28.9.06"
+ "text": "â“¢Laulont â“•Francois â“‘8"
},
"train-page_1-line_2": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_1-line_2.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_1-line_2.jpg",
"polygon": [
[
- 28,
- 328
+ 0,
+ 0
],
[
- 28,
- 465
+ 0,
+ 84
],
[
- 755,
- 465
+ 1900,
+ 84
],
[
- 755,
- 328
+ 1900,
+ 0
],
[
- 28,
- 328
+ 0,
+ 0
]
]
},
- "text": "â“¢Reboul â“•Jean â“‘30.9.02"
+ "text": "â“¢Ciret â“•Antoine â“‘27"
},
"train-page_1-line_3": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_1-line_3.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_1-line_3.jpg",
"polygon": [
[
- 23,
- 463
+ 0,
+ 0
],
[
- 23,
- 604
+ 0,
+ 85
],
[
- 803,
- 604
+ 1900,
+ 85
],
[
- 803,
- 463
+ 1900,
+ 0
],
[
- 23,
- 463
+ 0,
+ 0
]
]
},
- "text": "â“¢Bareyre â“•Jean â“‘28.3.11"
+ "text": "â“¢Ciret â“•Marie â“‘28"
},
"train-page_1-line_4": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_1-line_4.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_1-line_4.jpg",
"polygon": [
[
- 21,
- 604
+ 0,
+ 0
],
[
- 21,
- 743
+ 0,
+ 84
],
[
- 812,
- 743
+ 1900,
+ 84
],
[
- 812,
- 604
+ 1900,
+ 0
],
[
- 21,
- 604
+ 0,
+ 0
]
]
},
- "text": "â“¢Roussy â“•Jean â“‘4.11.14"
+ "text": "â“¢Ciret â“•Marie â“‘2"
},
"train-page_2-line_1": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_2-line_1.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_2-line_1.jpg",
"polygon": [
[
- 18,
- 197
+ 0,
+ 0
],
[
- 18,
- 340
+ 0,
+ 83
],
[
- 751,
- 340
+ 1900,
+ 83
],
[
- 751,
- 197
+ 1900,
+ 0
],
[
- 18,
- 197
+ 0,
+ 0
]
]
},
- "text": "â“¢Marin â“•Marcel â“‘10.8.06"
+ "text": "â“¢Eureston â“•Solange â“‘10"
},
"train-page_2-line_2": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_2-line_2.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_2-line_2.jpg",
"polygon": [
[
- 18,
- 340
+ 0,
+ 0
],
[
- 18,
- 476
+ 0,
+ 83
],
[
- 751,
- 476
+ 1900,
+ 83
],
[
- 751,
- 340
+ 1900,
+ 0
],
[
- 18,
- 340
+ 0,
+ 0
]
]
},
- "text": "â“¢Amical â“•Eloi â“‘11.10.04"
+ "text": "â“¢Terontussieux â“•Jean â“‘2"
},
"train-page_2-line_3": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/train-page_2-line_3.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/train-page_2-line_3.jpg",
"polygon": [
[
- 21,
- 476
+ 0,
+ 0
],
[
- 21,
- 615
+ 0,
+ 84
],
[
- 746,
- 615
+ 1900,
+ 84
],
[
- 746,
- 476
+ 1900,
+ 0
],
[
- 21,
- 476
+ 0,
+ 0
]
]
},
- "text": "â“¢Biros â“•Mael â“‘30.10.10"
+ "text": "â“¢Pressonet â“•Marie â“‘12"
}
},
"val": {
"val-page_1-line_1": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/val-page_1-line_1.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/val-page_1-line_1.jpg",
"polygon": [
[
- 14,
- 211
+ 0,
+ 0
],
[
- 14,
- 347
+ 0,
+ 84
],
[
- 755,
- 347
+ 1900,
+ 84
],
[
- 755,
- 211
+ 1900,
+ 0
],
[
- 14,
- 211
+ 0,
+ 0
]
]
},
- "text": "â“¢Monar⇠ⓕBouis â“‘29â‡â‡â‡04"
+ "text": "â“¢Cirau⇠ⓕAntoine â“‘â‡â‡"
},
"val-page_1-line_2": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/val-page_1-line_2.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/val-page_1-line_2.jpg",
"polygon": [
[
- 14,
- 350
+ 0,
+ 0
],
[
- 14,
- 484
+ 0,
+ 84
],
[
- 748,
- 484
+ 1900,
+ 84
],
[
- 748,
- 350
+ 1900,
+ 0
],
[
- 14,
- 350
+ 0,
+ 0
]
]
},
- "text": "â“¢Astier â“•Artâ‡ur â“‘11â‡2â‡13"
+ "text": "â“¢Cirau⇠ⓕPriser â“‘â‡â‡"
},
"val-page_1-line_3": {
"dataset_id": "dataset_id",
"image": {
- "iiif_url": "{FIXTURES}/extraction/images/text_line/val-page_1-line_3.jpg",
+ "iiif_url": "{FIXTURES}/extraction/images/val-page_1-line_3.jpg",
"polygon": [
[
- 11,
- 484
+ 0,
+ 0
],
[
- 11,
- 622
+ 0,
+ 84
],
[
- 751,
- 622
+ 1900,
+ 84
],
[
- 751,
- 484
+ 1900,
+ 0
],
[
- 11,
- 484
+ 0,
+ 0
]
]
},
- "text": "â“¢â‡e â‡lieâ‡er â“•Jules â“‘21â‡11â‡11"
+ "text": "â“¢Cirau⇠ⓕElisaâ‡et⇠ⓑâ‡â‡"
}
}
}
diff --git a/tests/data/extraction/tokens.yml b/tests/data/extraction/tokens.yml
index 23c24db674bb882f940b8e995e3a9cac75c8b231..c8b75e54ca249b10d9d31029652b82531b09d40b 100644
--- a/tests/data/extraction/tokens.yml
+++ b/tests/data/extraction/tokens.yml
@@ -5,6 +5,6 @@ surname:
firstname:
start: "â“•"
end: ""
-birthdate:
+age:
start: "â“‘"
end: ""
diff --git a/tests/data/tokens/end_tokens.yml b/tests/data/tokens/end_tokens.yml
index c660b7788fc5fafd48d197be29a55b5dd10e21fa..5f42b7e94b6b0942d2fa46884662c50531239db2 100644
--- a/tests/data/tokens/end_tokens.yml
+++ b/tests/data/tokens/end_tokens.yml
@@ -11,7 +11,7 @@ Person /!\:
adj:
start: â’¼
end: â’½
-birthdate:
+age:
start: â’¾
end: â’¿
firstname:
diff --git a/tests/data/tokens/no_end_tokens.yml b/tests/data/tokens/no_end_tokens.yml
index a19597267d1952aad03b58c66dafa5fe43e5f52b..f840973a099749cc663612bb6f9c407e73104a59 100644
--- a/tests/data/tokens/no_end_tokens.yml
+++ b/tests/data/tokens/no_end_tokens.yml
@@ -11,7 +11,7 @@ Person /!\:
adj:
start: â’¹
end: ''
-birthdate:
+age:
start: â’º
end: ''
firstname:
diff --git a/tests/test_db.py b/tests/test_db.py
index 875add45c44d3e395296c5c9f0f5aad90fdf38e4..088cab36cf3d07aa1546e1f0ef49f14d904a7d51 100644
--- a/tests/test_db.py
+++ b/tests/test_db.py
@@ -84,7 +84,7 @@ def test_get_transcriptions(sources, mock_database):
if not sources or False in sources:
expected_transcriptions.append(
{
- "text": "Caillet Maurice 28.9.06",
+ "text": "Laulont Francois 8",
"worker_version_id": None,
"worker_run_id": None,
}
@@ -93,7 +93,7 @@ def test_get_transcriptions(sources, mock_database):
if not sources or "id" in sources:
expected_transcriptions.append(
{
- "text": "caillet maurice 28.9.06",
+ "text": "laulont francois 8",
"worker_version_id": "worker_version_id",
"worker_run_id": "worker_run_id",
}
@@ -121,7 +121,7 @@ def test_get_transcriptions(sources, mock_database):
@pytest.mark.parametrize("source", (False, "id", None))
@pytest.mark.parametrize(
- "supported_types", (["surname"], ["surname", "firstname", "birthdate"])
+ "supported_types", (["surname"], ["surname", "firstname", "age"])
)
def test_get_transcription_entities(source, mock_database, supported_types):
worker_version = f"worker_version_{source}" if isinstance(source, str) else source
@@ -137,22 +137,22 @@ def test_get_transcription_entities(source, mock_database, supported_types):
expected_entities = [
{
- "name": "Caillet",
+ "name": "Laulont",
"type": "surname",
"offset": 0,
"length": 7,
},
{
- "name": "Maurice",
+ "name": "Francois",
"type": "firstname",
"offset": 9,
- "length": 7,
+ "length": 8,
},
{
- "name": "28.9.06",
- "type": "birthdate",
- "offset": 18,
- "length": 7,
+ "name": "8",
+ "type": "age",
+ "offset": 19,
+ "length": 1,
},
]
diff --git a/tests/test_download.py b/tests/test_download.py
index 8dd39fc49397e3b0a8bf0cb2a0f528ecfc06e498..1c1879fb56ef6d664564ccf5e11bfc264385452b 100644
--- a/tests/test_download.py
+++ b/tests/test_download.py
@@ -93,26 +93,26 @@ def test_download(split_content, monkeypatch, tmp_path):
# Check "labels.json"
expected_labels = {
"test": {
- "images/test/dataset_id/test-page_1-line_1.jpg": "â“¢Couâ‡e⇠ⓕBouis â“‘â‡.12.14",
- "images/test/dataset_id/test-page_1-line_2.jpg": "â“¢â‡outrain â“•Aâ‡olâ‡â‡e â“‘9.4.13",
- "images/test/dataset_id/test-page_1-line_3.jpg": "â“¢â‡abale â“•â‡ranâ‡ais â“‘26.3.11",
- "images/test/dataset_id/test-page_2-line_1.jpg": "â“¢â‡urosoy â“•Bouis â“‘22â‡4â‡18",
- "images/test/dataset_id/test-page_2-line_2.jpg": "â“¢Colaiani â“•Anâ‡els â“‘28.11.1â‡",
- "images/test/dataset_id/test-page_2-line_3.jpg": "â“¢Renouar⇠ⓕMaurice â“‘2â‡.â‡.04",
+ "images/test/dataset_id/test-page_1-line_1.jpg": "â“¢Leunaut â“•Clauâ‡e â“‘â‡â‡",
+ "images/test/dataset_id/test-page_1-line_2.jpg": "â“¢â‡auracâ‡o â“•Clauâ‡ine â“‘â‡â‡",
+ "images/test/dataset_id/test-page_1-line_3.jpg": "â“¢Laurent â“•Jacâ‡use â“‘21",
+ "images/test/dataset_id/test-page_2-line_1.jpg": "â“¢â‡alette â“•Elisaâ‡et⇠ⓑ7â‡",
+ "images/test/dataset_id/test-page_2-line_2.jpg": "â“¢Tanâ‡ol â“•Jean â“‘7â‡",
+ "images/test/dataset_id/test-page_2-line_3.jpg": "â“¢â‡auret â“•Jean â“‘â‡â‡",
},
"train": {
- "images/train/dataset_id/train-page_1-line_1.jpg": "â“¢Caillet â“•Maurice â“‘28.9.06",
- "images/train/dataset_id/train-page_1-line_2.jpg": "â“¢Reboul â“•Jean â“‘30.9.02",
- "images/train/dataset_id/train-page_1-line_3.jpg": "â“¢Bareyre â“•Jean â“‘28.3.11",
- "images/train/dataset_id/train-page_1-line_4.jpg": "â“¢Roussy â“•Jean â“‘4.11.14",
- "images/train/dataset_id/train-page_2-line_1.jpg": "â“¢Marin â“•Marcel â“‘10.8.06",
- "images/train/dataset_id/train-page_2-line_2.jpg": "â“¢Amical â“•Eloi â“‘11.10.04",
- "images/train/dataset_id/train-page_2-line_3.jpg": "â“¢Biros â“•Mael â“‘30.10.10",
+ "images/train/dataset_id/train-page_1-line_1.jpg": "â“¢Laulont â“•Francois â“‘8",
+ "images/train/dataset_id/train-page_1-line_2.jpg": "â“¢Ciret â“•Antoine â“‘27",
+ "images/train/dataset_id/train-page_1-line_3.jpg": "â“¢Ciret â“•Marie â“‘28",
+ "images/train/dataset_id/train-page_1-line_4.jpg": "â“¢Ciret â“•Marie â“‘2",
+ "images/train/dataset_id/train-page_2-line_1.jpg": "â“¢Eureston â“•Solange â“‘10",
+ "images/train/dataset_id/train-page_2-line_2.jpg": "â“¢Terontussieux â“•Jean â“‘2",
+ "images/train/dataset_id/train-page_2-line_3.jpg": "â“¢Pressonet â“•Marie â“‘12",
},
"val": {
- "images/val/dataset_id/val-page_1-line_1.jpg": "â“¢Monar⇠ⓕBouis â“‘29â‡â‡â‡04",
- "images/val/dataset_id/val-page_1-line_2.jpg": "â“¢Astier â“•Artâ‡ur â“‘11â‡2â‡13",
- "images/val/dataset_id/val-page_1-line_3.jpg": "â“¢â‡e â‡lieâ‡er â“•Jules â“‘21â‡11â‡11",
+ "images/val/dataset_id/val-page_1-line_1.jpg": "â“¢Cirau⇠ⓕAntoine â“‘â‡â‡",
+ "images/val/dataset_id/val-page_1-line_2.jpg": "â“¢Cirau⇠ⓕPriser â“‘â‡â‡",
+ "images/val/dataset_id/val-page_1-line_3.jpg": "â“¢Cirau⇠ⓕElisaâ‡et⇠ⓑâ‡â‡",
},
}
@@ -124,9 +124,7 @@ def test_download(split_content, monkeypatch, tmp_path):
continue
assert ImageChops.difference(
- Image.open(
- EXTRACTION_DATA_PATH / "images" / "text_line" / expected_path.name
- ),
+ Image.open(EXTRACTION_DATA_PATH / "images" / expected_path.name),
Image.open(expected_path),
)
diff --git a/tests/test_extract.py b/tests/test_extract.py
index 368d99984dace2f1de58708b8fc06c340e6151c6..922d1026fe95d0f47b09b748fad5e3202ef10e3c 100644
--- a/tests/test_extract.py
+++ b/tests/test_extract.py
@@ -114,117 +114,117 @@ def test_process_element_unknown_token_in_text_error(mock_database, tmp_path):
True,
True,
"worker_version_id",
- """â– â“¢ c a i l l e t â– â“• m a u r i c e â– â“‘ 28. 9.0 6
-â– â“¢ re b ou l â– â“• j e a n â– â“‘ 30. 9.0 2
-â– â“¢ b a re y re â– â“• j e a n â– â“‘ 28. 3 . 1 1
-â– â“¢ r ou s s y â– â“• j e a n â– â“‘ 4 . 1 1 . 1 4
-â– â“¢ m a r i n â– â“• m a r c e l â– â“‘ 1 0 . 8 . 0 6
-â– â“¢ a m i c a l â– â“• e l o i â– â“‘ 1 1 . 1 0 . 0 4
-â– â“¢ b i r o s â– â“• m a e l â– â“‘ 30. 1 0 . 1 0""",
+ """â– â“¢ l a u l ont â– â“• f r an c oi s â– â“‘ 8
+â– â“¢ c i re t â– â“• an t oi ne â– â“‘ 2 7
+â– â“¢ c i re t â– â“• m a r ie â– â“‘ 2 8
+â– â“¢ c i re t â– â“• m a r ie â– â“‘ 2
+â– â“¢ e u re s t on â– â“• so l an g e â– â“‘ 1 0
+â– â“¢ t e r ont u s s ie u x â– â“• j e an â– â“‘ 2
+â– â“¢ p re s s on e t â– â“• m a r ie â– â“‘ 1 2""",
40,
),
(
True,
False,
"worker_version_id",
- """â– â“¢ c a i l l e t â– â“• m a u r i c e â– â“‘ 28. 9.0 6
-â– â“¢ re b ou l â– â“• j e a n â– â“‘ 30. 9.0 2
-â– â“¢ b a re y re â– â“• j e a n â– â“‘ 28. 3 . 1 1
-â– â“¢ r ou s s y â– â“• j e a n â– â“‘ 4 . 1 1 . 1 4
-â– â“¢ m a r i n â– â“• m a r c e l â– â“‘ 1 0 . 8 . 0 6
-â– â“¢ a m i c a l â– â“• e l o i â– â“‘ 1 1 . 1 0 . 0 4
-â– â“¢ b i r o s â– â“• m a e l â– â“‘ 30. 1 0 . 1 0""",
+ """â– â“¢ l a u l ont â– â“• f r an c oi s â– â“‘ 8
+â– â“¢ c i re t â– â“• an t oi ne â– â“‘ 2 7
+â– â“¢ c i re t â– â“• m a r ie â– â“‘ 2 8
+â– â“¢ c i re t â– â“• m a r ie â– â“‘ 2
+â– â“¢ e u re s t on â– â“• so l an g e â– â“‘ 1 0
+â– â“¢ t e r ont u s s ie u x â– â“• j e an â– â“‘ 2
+â– â“¢ p re s s on e t â– â“• m a r ie â– â“‘ 1 2""",
40,
),
(
False,
True,
"worker_version_id",
- """â– ca i l l e t â– ma u r i ce â– 28. 9.0 6
-â– re b o u l â– j e a n â– 30. 9.0 2
-â– b a re y re â– j e a n â– 28. 3 . 1 1
-â– r o u s s y â– j e a n â– 4 . 11.1 4
-â– ma r i n â– ma r ce l â– 10. 8 . 0 6
-â– a m i ca l â– el o i â– 11.1 0 . 0 4
-â– b i r o s â– ma el â– 30. 10. 1 0""",
+ """â– la u l ont â– f r an c oi s â– 8
+â– c i re t â– an t oi ne â– 2 7
+â– c i re t â– m a r ie â– 2 8
+â– c i re t â– m a r ie â– 2
+â– e u res t on â– so l an g e â– 1 0
+â– t e r ont u ss ie u x â– j e an â– 2
+â– p res so ne t â– m a r ie â– 1 2""",
40,
),
(
False,
False,
"worker_version_id",
- """â– ca i l l e t â– ma u r i ce â– 28. 9.0 6
-â– re b o u l â– j e a n â– 30. 9.0 2
-â– b a re y re â– j e a n â– 28. 3 . 1 1
-â– r o u s s y â– j e a n â– 4 . 11.1 4
-â– ma r i n â– ma r ce l â– 10. 8 . 0 6
-â– a m i ca l â– el o i â– 11.1 0 . 0 4
-â– b i r o s â– ma el â– 30. 10. 1 0""",
+ """â– la u l ont â– f r an c oi s â– 8
+â– c i re t â– an t oi ne â– 2 7
+â– c i re t â– m a r ie â– 2 8
+â– c i re t â– m a r ie â– 2
+â– e u res t on â– so l an g e â– 1 0
+â– t e r ont u ss ie u x â– j e an â– 2
+â– p res so ne t â– m a r ie â– 1 2""",
40,
),
(
True,
True,
False,
- """â– â“¢ C a i l l e t â– â“• M a u r i c e â– â“‘ 2 8 . 9 . 0 6
-â– â“¢ R e b o u l â– â“• J e a n â– â“‘ 3 0 . 9 . 0 2
-â– â“¢ B a r e y r e â– â“• J e a n â– â“‘ 2 8 . 3 . 1 1
-â– â“¢ R o u s s y â– â“• J e a n â– â“‘ 4 . 1 1 . 1 4
-â– â“¢ M a r i n â– â“• M a r c e l â– â“‘ 1 0 . 8 . 0 6
-â– â“¢ A m i c a l â– â“• E l o i â– â“‘ 1 1 . 1 0 . 0 4
-â– â“¢ B i r o s â– â“• M a e l â– â“‘ 3 0 . 1 0 . 1 0""",
+ """â– â“¢ L a u l o n t â– â“• F r a n c o i s â– â“‘ 8
+â– â“¢ C i r e t â– â“• A n t o i n e â– â“‘ 2 7
+â– â“¢ C i r e t â– â“• M a r ie â– â“‘ 2 8
+â– â“¢ C i r e t â– â“• M a r ie â– â“‘ 2
+â– â“¢ E u r e s t o n â– â“• S o l a n g e â– â“‘ 1 0
+â– â“¢ T e r o n t u s s ie u x â– â“• J e a n â– â“‘ 2
+â– â“¢ P r e s s o n e t â– â“• M a r ie â– â“‘ 1 2""",
40,
),
(
True,
True,
False,
- """â– â“¢ C a i l l e t â– â“• M a u ri ce â– â“‘ 28. 9.0 6
-â– â“¢ R e b ou l â– â“• J e a n â– â“‘ 30. 9.0 2
-â– â“¢ B a re y re â– â“• J e a n â– â“‘ 28. 3 . 1 1
-â– â“¢ R ou s s y â– â“• J e a n â– â“‘ 4 . 11.1 4
-â– â“¢ Mar i n â– â“• Mar ce l â– â“‘ 10. 8 . 0 6
-â– â“¢ A m ic a l â– â“• E l o i â– â“‘ 11.1 0 . 0 4
-â– â“¢ B i r o s â– â“• M a e l â– â“‘ 30. 10. 10""",
- 55,
+ """â– â“¢ L a u l ont â– â“• F r an c oi s â– â“‘ 8
+â– â“¢ C i re t â– â“• A n t oi n e â– â“‘ 2 7
+â– â“¢ C i re t â– â“• M a r ie â– â“‘ 2 8
+â– â“¢ C i re t â– â“• M a r ie â– â“‘ 2
+â– â“¢ E u re s t on â– â“• S o l an g e â– â“‘ 1 0
+â– â“¢ T e r ont u s s ie u x â– â“• J e an â– â“‘ 2
+â– â“¢ P re s s on e t â– â“• M a r ie â– â“‘ 1 2""",
+ 45,
),
(
True,
False,
False,
- """â– â“¢ C a i l l e t â– â“• M a u r i c e â– â“‘ 2 8 . 9 . 0 6
-â– â“¢ R e b o u l â– â“• J e a n â– â“‘ 3 0 . 9 . 0 2
-â– â“¢ B a r e y r e â– â“• J e a n â– â“‘ 2 8 . 3 . 1 1
-â– â“¢ R o u s s y â– â“• J e a n â– â“‘ 4 . 1 1 . 1 4
-â– â“¢ M a r i n â– â“• M a r c e l â– â“‘ 1 0 . 8 . 0 6
-â– â“¢ A m i c a l â– â“• E l o i â– â“‘ 1 1 . 1 0 . 0 4
-â– â“¢ B i r o s â– â“• M a e l â– â“‘ 3 0 . 1 0 . 1 0""",
+ """â– â“¢ L a u l o n t â– â“• F r a n c o i s â– â“‘ 8
+â– â“¢ C i r e t â– â“• A n t o i n e â– â“‘ 2 7
+â– â“¢ C i r e t â– â“• M a r ie â– â“‘ 2 8
+â– â“¢ C i r e t â– â“• M a r ie â– â“‘ 2
+â– â“¢ E u r e s t o n â– â“• S o l a n g e â– â“‘ 1 0
+â– â“¢ T e r o n t u s s ie u x â– â“• J e a n â– â“‘ 2
+â– â“¢ P r e s s o n e t â– â“• M a r ie â– â“‘ 1 2""",
40,
),
(
False,
True,
False,
- """â– C a i l l e t â– Ma u r i c e â– 28. 9.0 6
-â– R e b o u l â– J e a n â– 30. 9.0 2
-â– B a r e y r e â– J e a n â– 28. 3 . 1 1
-â– R o u s s y â– J e a n â– 4 . 1 1 . 1 4
-â– Ma r i n â– Ma r c e l â– 1 0 . 8 . 0 6
-â– A m i c a l â– E l o i â– 1 1 . 1 0 . 0 4
-â– B i r o s â– Ma e l â– 30. 1 0 . 1 0""",
+ """â– L a u l ont â– F r an c oi s â– 8
+â– C i re t â– A n t oi n e â– 2 7
+â– C i re t â– M a r ie â– 2 8
+â– C i re t â– M a r ie â– 2
+â– E u re s t on â– S o l an g e â– 1 0
+â– T e r ont u s s ie u x â– J e an â– 2
+â– P re s s on e t â– M a r ie â– 1 2""",
40,
),
(
False,
False,
False,
- """â– C a i l l e t â– Ma u r i c e â– 28. 9.0 6
-â– R e b o u l â– J e a n â– 30. 9.0 2
-â– B a r e y r e â– J e a n â– 28. 3 . 1 1
-â– R o u s s y â– J e a n â– 4 . 1 1 . 1 4
-â– Ma r i n â– Ma r c e l â– 1 0 . 8 . 0 6
-â– A m i c a l â– E l o i â– 1 1 . 1 0 . 0 4
-â– B i r o s â– Ma e l â– 30. 1 0 . 1 0""",
+ """â– L a u l ont â– F r an c oi s â– 8
+â– C i re t â– A n t oi n e â– 2 7
+â– C i re t â– M a r ie â– 2 8
+â– C i re t â– M a r ie â– 2
+â– E u re s t on â– S o l an g e â– 1 0
+â– T e r ont u s s ie u x â– J e an â– 2
+â– P re s s on e t â– M a r ie â– 1 2""",
40,
),
),
@@ -366,21 +366,21 @@ def test_extract(
assert set(pickle.loads((output / "charset.pkl").read_bytes())) == expected_charset
# Check "language_corpus.txt"
- expected_char_language_corpus = """â“¢ C a i l l e t â– â– â“• M a u r i c e â– â– â“‘ 2 8 . 9 . 0 6
-â“¢ R e b o u l â– â– â“• J e a n â– â– â“‘ 3 0 . 9 . 0 2
-â“¢ B a r e y r e â– â– â“• J e a n â– â– â“‘ 2 8 . 3 . 1 1
-â“¢ R o u s s y â– â– â“• J e a n â– â– â“‘ 4 . 1 1 . 1 4
-â“¢ M a r i n â– â– â“• M a r c e l â– â– â“‘ 1 0 . 8 . 0 6
-â“¢ A m i c a l â– â– â“• E l o i â– â– â“‘ 1 1 . 1 0 . 0 4
-â“¢ B i r o s â– â– â“• M a e l â– â– â“‘ 3 0 . 1 0 . 1 0"""
-
- expected_word_language_corpus = """â“¢ Caillet â– â“• Maurice â– â“‘ 28 â– . â– 9 â– . â– 06
-â“¢ Reboul â– â“• Jean â– â“‘ 30 â– . â– 9 â– . â– 02
-â“¢ Bareyre â– â“• Jean â– â“‘ 28 â– . â– 3 â– . â– 11
-â“¢ Roussy â– â“• Jean â– â“‘ 4 â– . â– 11 â– . â– 14
-â“¢ Marin â– â“• Marcel â– â“‘ 10 â– . â– 8 â– . â– 06
-â“¢ Amical â– â“• Eloi â– â“‘ 11 â– . â– 10 â– . â– 04
-â“¢ Biros â– â“• Mael â– â“‘ 30 â– . â– 10 â– . â– 10"""
+ expected_char_language_corpus = """â“¢ L a u l o n t â– â– â“• F r a n c o i s â– â– â“‘ 8
+â“¢ C i r e t â– â– â“• A n t o i n e â– â– â“‘ 2 7
+â“¢ C i r e t â– â– â“• M a r i e â– â– â“‘ 2 8
+â“¢ C i r e t â– â– â“• M a r i e â– â– â“‘ 2
+â“¢ E u r e s t o n â– â– â“• S o l a n g e â– â– â“‘ 1 0
+â“¢ T e r o n t u s s i e u x â– â– â“• J e a n â– â– â“‘ 2
+â“¢ P r e s s o n e t â– â– â“• M a r i e â– â– â“‘ 1 2"""
+
+ expected_word_language_corpus = """â“¢ Laulont â– â“• Francois â– â“‘ 8
+â“¢ Ciret â– â“• Antoine â– â“‘ 27
+â“¢ Ciret â– â“• Marie â– â“‘ 28
+â“¢ Ciret â– â“• Marie â– â“‘ 2
+â“¢ Eureston â– â“• Solange â– â“‘ 10
+â“¢ Terontussieux â– â“• Jean â– â“‘ 2
+â“¢ Pressonet â– â“• Marie â– â“‘ 12"""
if existing:
expected_char_language_corpus = (
@@ -511,7 +511,7 @@ def test_extract_transcription_no_translation(mock_database, tokens, tmp_path):
).execute()
# Early return with only the element transcription text instead of a translation
- assert extractor.extract_transcription(element) == "Coupez Bouis 7.12.14"
+ assert extractor.extract_transcription(element) == "Leunaut Claude 49"
@pytest.mark.parametrize(