IGN Train and Validation Data for ICDAR'25 MapText Competition

Data set of 2Kx2K image tiles cropped from Napoleonic Cadastre maps of the Val de Marne Archive for the ICDAR'25 Competition on Historical Map Text Detection, Recognition, and Linking.

Annotations and images follow the format described at the competition website and can be evaluated using the official evaluation repository script.

This dataset is a superset of the dataset used in the 2024 edition: the first 80 image from the training set, and the first 15 images from the validation set are the same as the version 1.1 of the IGN Train and Validation Data for ICDAR'24 MapText Competition. However, minor issues in their annotations may have been fixed, so you should use this new dataset instead.

Please note the we also provide an extra synthetic dataset for training, which is released under a different record: "IGN Synthetic Train Data for ICDAR'25 MapText Competition" (10.5281/zenodo.14394546).

Train
Validation

Annotations
ign25_train.json
ign25_val.json

Images
train.zip
val.zip

Files
ign25/train/*.jpg
ign25/val/*.jpg

Tiles
228
25

Map Sheets
78
12

Words
25,564
2,725

Label Groups
23,542
2,413

Illegible Words
1,684
274

Truncated Words
1,351
129

Valid Words
23,880
2,451

Original images available at https://archives.valdemarne.fr/recherches/archives-en-ligne/cadastre-napoleonien as of 11 Dec. 2024.