README.md

---
    datasets:
    - imagefolder
    pretty_name: LM2rulers38squarified
    task_categories:
    - image-classification
    language:
    - und
    size_categories:
    - n<10607
    ---

    # LM2rulers38squarified

    Squarified (720px) variant of LM2rulers38. Images organized by ruler_class, split into train/val/test.

    ## Structure

    This repository uses the **imagefolder** format:

    LM2rulers38squarified/
    train/
    <ruler_class>/.jpg
    val/
    <ruler_class>/.jpg
    test/
    <ruler_class>/*.jpg
    - Classes with fewer than 10 images were **skipped**.
    - For all included classes, an **80/10/10** split is created with at least **1 image per split per class** (deterministic shuffling).

    ## Stats

    - Classes included: **20**
    - Images: train **8478**, val **1052**, test **1077**

    <details>
    <summary>Skipped classes (fewer than 10 images)</summary>

    None
    </details>

    ## Loading

    ```python
    from datasets import load_dataset

    ds = load_dataset("phyloforfun/LM2rulers38squarified")
    print(ds)