Unpaired Horse and Zebra Images for CycleGAN Translation

A collection of two unpaired sets of photos of horses and zebras, designed for unpaired image-to-image translation as introduced in the CycleGAN paper. The dataset is categorized under the '1K<n<10K' size range, indicating it contains between 1,000 and 10,000 images. It was created by 'gigant' and last updated on October 24, 2022.

Use Cases

• Train a CycleGAN model to perform unpaired image-to-image translation between horse and zebra domains.
• Benchmark generative adversarial network architectures using the distinct visual features of horses and zebras.
• Study domain adaptation techniques on the unpaired sets of horse and zebra photos.

Strengths

• Dataset size is categorized as '1K<n<10K', indicating a moderate scale suitable for model training.
• Specifically designed for the unpaired image translation task as presented in the seminal CycleGAN paper (arXiv:1703.10593).
• Last update recorded on October 24, 2022, providing a timestamp for the hosted version.

Limitations

• The exact number of images, file formats, and specific image attributes are unknown.
• The dataset's geographic origin is only broadly indicated as 'Region:us', which may limit geographic diversity.
• As an unpaired dataset, it lacks corresponding image pairs, which restricts its use for supervised translation tasks.

Provenance

Source

huggingface

Collection Method

Two unpaired sets of photos collected for the purpose of unpaired image-to-image translation.

Freshness

2022-10-24

Geography

Region is tagged as 'us'.