Unpaired Image Collection for CycleGAN Research

Serving as part of the CycleGAN datasets collection, originally hosted by researchers at UC Berkeley. It is used for unpaired image-to-image translation tasks as described in the associated 2017 arXiv paper. The dataset contains images, but the specific row count, column structure, and file formats are unknown.

Use Cases

• Train Cycle-Consistent Adversarial Networks for unpaired image-to-image translation using the provided image collections.
• Benchmark image translation models on a dataset associated with the foundational CycleGAN research paper (arXiv:1703.10593).
• Apply computer vision techniques to the image data, which is categorized under the 'image' modality tag.

Strengths

• Associated with a seminal research paper on CycleGANs (arXiv:1703.10593), providing a standard benchmark.
• Hosted on Hugging Face with tags indicating support for multiple data libraries (pandas, polars, datasets).
• Dataset was last updated on the Hugging Face platform in April 2022.

Limitations

• The dataset size, number of images, and specific file formats are not provided in the input.
• No sample data or column descriptions are available to assess content or structure.
• The temporal and geographic coverage of the images is unknown.

Provenance

Source

Originally hosted by researchers at UC Berkeley as part of the CycleGAN datasets.

Collection Method

null

Time Range

null

Freshness

null

Geography

null