Banana100: Image Degradation from 100 Iterative AI Edits

A dataset demonstrating quality degradation caused by 100 iterative edits using the Nano Banana Pro model. It was created by kenantang and last updated on April 10, 2026. The dataset illustrates how imperceptible noise from single-step edits compounds over repeated applications, eventually degrading images to an unusable level.

Use Cases

• Quantify cumulative noise in generative models based on the described iterative editing process.
• Benchmark image quality degradation over multiple editing steps based on the described failure case.
• Study the robustness of image generation pipelines based on the described compounding noise phenomenon.

Strengths

• Focuses on a specific, documented failure case of iterative AI editing.
• Created using a state-of-the-art model (Nano Banana Pro) as mentioned in the description.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Likely generated by applying the Nano Banana Pro model for iterative edits.

Time Range

null

Freshness

Last updated 2026-04-10 16:19:35; freshness should be verified.

Geography

null