Denoised s-SNOM Infrared Microscopy Images via CycleGAN

A collection of scattering-type scanning near-field optical microscopy (s-SNOM) images used to train a machine-learning denoising model. The data was generated by Carlos Baiz and last updated on October 15, 2025. The dataset likely contains paired or unpaired sets of images captured at rapid and extended acquisition times.

Use Cases

• Training denoising models for s-SNOM images based on the described patch-based machine learning reconstruction method.
• Benchmarking image quality enhancement techniques based on structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR) metrics mentioned in the description.
• Reducing data acquisition time in point-scanning probe microscopy based on the method's potential to match quality from longer integration times.
• Preserving topographical and molecular information in weakly absorbing samples based on the described challenge and solution.

Strengths

• Method demonstrates significant image quality improvements, with results comparable to images captured with four times the integration time.
• The machine learning approach allows for flexible reconstruction of images of arbitrary sizes, addressing a critical need in point-scanning microscopy.
• Model is trained on unpaired image sets, which can simplify data collection requirements.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset's specific file formats and size are not provided in the metadata.

Provenance

Source

Texas Data Repository Harvested Dataverse

Collection Method

Images captured via scattering-type scanning near-field optical microscopy (s-SNOM).

Time Range

null

Freshness

Last updated 2025-10-15 07:02:01; freshness should be verified.

Geography

null