About the Dataset Format

[Jll0716]

I want to know what the meaning of dividing the dataset into these formats is. Is it categorized by animal species? When I train, can I choose to use only split1 or 2\3? What impact will this have？
│ |-- ap10k-train-split2.json
│ |-- ap10k-train-split3.json
│ │-- ap10k-val-split1.json
│ |-- ap10k-val-split2.json
│ |-- ap10k-val-split3.json
│ |-- ap10k-test-split1.json
│ |-- ap10k-test-split2.json
│ |-- ap10k-test-split3.json