The Czech Traveler dataset is based on a collection of 1,276 images taken by a professional photographer during trips to Albania, Corsica, Romania, Slovenia and Ukraine. These images have short textual annotations consisting of 1 to 10 words saved in images' EXIF data. Out of the annotations we extracted 103 unique annotations. The image annotations were broken into entities and these entities were assigned a label (class).
| Key metrics | download | |
|---|---|---|
| all images | 1,276 | zip (51.1 MB) |
| unique annotations | 103 | |
| entities | 186 | |
| labeled entities | 184 | |
| labeled entities (to 9 classes) with inter-annotator agreement | 143 | ods (25 KB) |
| unique entities | 151 | |
| named entities | 101 | |
| unique named entities | 76 | |
| unique entities with inter-annotator agreement | 113 | |
| entities for which not even the head was mapped to WordNet | 47 | |
| unique entities for which not even the head was mapped to WordNet | 41 | |
| entities for which not even the head was mapped to WordNet among the 143 entities with inter-annotator agreemen | 30 | |