← All notes

Short answer

Semantic segmentation pairs each image with a mask that labels every pixel. Apply the same geometric augmentation to image and mask together (Albumentations takes both), and run it on the fly with set_transform(). [S1][S2]

Claims

• id: C1 claim: Segmentation datasets classify every pixel in an image. source: S1 evidence: "Semantic segmentation datasets are used to train a model to classify every pixel in an"
• id: C2 claim: The annotation is a segmentation mask of the image. source: S1 evidence: "annotation: segmentation mask of the image."
• id: C3 claim: Augmentations are applied with set_transform(). source: S2 evidence: "Apply data augmentations to a dataset with set_transform()."

Sources

• [S1] Semantic segmentation · Hugging Face. Hugging Face (datasets docs). https://huggingface.co/docs/datasets/semantic_segmentation . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-04.
• [S2] Process image data · Hugging Face. Hugging Face (datasets docs). https://huggingface.co/docs/datasets/image_process . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-04.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).

@misc{iil_note_q8r48v87,
  author = {{iinventedlight}},
  title = {{Prepare semantic segmentation datasets}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-04},
  url = {https://iinventedlight.com/library/notes/hf-datasets-semantic-segmentation/},
  note = {Permanent ID iil:note:q8r48v87. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt. This page in Markdown: /library/notes/hf-datasets-semantic-segmentation.md.