DAMNDEVICE-AGNOSTIC SUPER-RESOLUTION

Improve your image.

From stars to molecules: DAMN reconstructs fine spatial detail from a single intensity image, using a neural network trained across a range of imaging conditions. Suitable for objects composed of (many) point-like sources such as stars, atoms, molecules, quantum dots, color centers, etc.

or drop it here · PNG, JPEG, WebP, single-frame TIFF · up to 200 × 200 px

Select a small scientific image or crop, up to 200 × 200 pixels and 20 MB.

Processed on your device

Loading the Andromeda example and its precomputed reconstruction…

01 Original input—

Your image
starts here

02 Super-resolved8× grid

DAMN
reconstruction

03 Gaussian-smoothedσ = 4 px

A clearer view
of the reconstruction

Inferno color map. Equal display size, different pixel grids. Each panel maps its own linear intensity range to color. PNG downloads contain grayscale reconstructions without a color palette at the full output resolution, linearly scaled and rounded to 8-bit intensities. Inputs are background-subtracted and normalized to a total intensity of width × height / 2500, matching the model's 50 × 50 training scale. The output grid is 8× finer in each dimension. Gaussian σ = 4 output pixels equals 0.5 input pixels.