Albedo

Per-pixel albedo = intrinsic diffuse color, with illumination factored out. Great for relighting and clothing color analysis.

Same input pipeline as segmentation and normals

All dense tasks share the same interface. Here's the input image used across all guide examples:

Sample input image

Signature

sapiens_albedo(
    input_path:  str,
    output_dir:  str,
    model_size:  str = "0.4b",
    device:      str = "cuda:0",
    save_pred:   bool = True,
) -> dict

Example

from strands_sapiens import sapiens_albedo

sapiens_albedo(
    input_path="person.jpg",
    output_dir="out/",
)

Output:

• out/person.jpg - side-by-side input vs. albedo viz (min-max normalized)
• out/person_albedo.npy - 3×H×W float array

Why this is useful

Photos mix three signals:

pixel_color ≈ albedo × shading × camera_response

Albedo gives you just the first factor. Once you have it you can:

• Relight (combine with normals).
• Color-match garments across photos under different lighting.
• Remove specular highlights (especially on skin / hair).

Visualization

The wrapper normalizes the predicted albedo to the [0, 255] range per image for display. The .npy keeps the original float values.

Tips

• Indoor vs. outdoor: the model generalizes remarkably well; lighting condition doesn't need to match training data.
• For relighting: pair albedo with normals and run a simple shading model (color = albedo × max(0, n · l)).

Related

• Normals → · Pointmap →

Albedo data flow