---
            pretty_name: Google Landmarks Dataset v2 (images only)
            license: unknown
            ---

            ## How to load

            ```python
            from datasets import load_dataset, Image

            ds = load_dataset("parquet", data_files="data/*.parquet", split="train")
            ds = ds.cast_column("image", Image())   # decode bytes -> PIL
            print(ds[0]["image"], ds[0]["image_id"])