A robust utility for automatically partitioning image datasets and annotations into training, validation, and testing subsets.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install dataset-splitter
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install dataset-splitter using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Dataset Splitter?
The Dataset Splitter is an essential tool designed to automate the data preparation phase of the machine learning lifecycle. It allows developers and data scientists to efficiently organize large volumes of raw images and corresponding metadata into structured directories. By integrating this tool into your workflow via Openclaw Skills, you can ensure that your datasets are perfectly balanced and formatted for various computer vision frameworks.
This skill eliminates the manual effort involved in file management by providing high-level commands for random and stratified splitting. Whether you are working on a small prototype or a production-grade model, this utility ensures your data is partitioned with precision and consistency.
Dataset Splitter Use Cases
- Preparing raw image folders for supervised learning by creating distinct training and evaluation pools.
- Generating YOLO-compliant directory structures for object detection projects.
- Balancing class distributions across datasets using stratified sampling to avoid model bias.
- Standardizing data pipelines for reproducible machine learning experiments using random seeds.
How Dataset Splitter Works
- The user specifies the source directory containing the images and, optionally, the associated annotations.
- The skill analyzes the input parameters such as split ratios (e.g., 80% train, 10% val, 10% test).
- If stratification is requested, the tool identifies class labels to maintain consistent distribution across all sets.
- The script partitions the files, ensuring that image-annotation pairs remain synchronized during the transfer.
- Files are moved or copied into a structured output directory, ready for immediate use in training scripts.
Dataset Splitter Setup
To get started with this tool from Openclaw Skills, first ensure you have the required dependencies installed:
pip install pillow
Basic usage to split a dataset with an 80/10/10 ratio:
python scripts/splitter.py split /path/to/images/ --ratios 80 10 10
Dataset Splitter Data Schema & Taxonomy
The Dataset Splitter organizes files into a clear hierarchy based on the split configuration:
| Component | Description |
|---|---|
/train |
Contains the majority of the data used for model optimization. |
/val |
Contains data used for hyperparameter tuning during training. |
/test |
Contains held-out data for final performance evaluation. |
--yolo flag |
Organizes the output into images/ and labels/ subdirectories for each split. |
Dataset Splitter Advanced Features
- Stratified Splitting: Ensures that each subset contains a representative percentage of every class.
- Annotation Support: Automatically detects and moves label files corresponding to each image.
- YOLO Format Export: Direct compatibility with popular object detection architectures.
- Reproducibility: Use the
--seedflag to generate identical splits across different environments. - Flexible Operations: Supports both copying files (preserving originals) and moving files for storage efficiency.
SKILL.md
Loading
Related Openclaw Skills
Daily Horoscope & Almanac
A specialized agent for querying traditional Chinese Almanac data, providing daily auspicious and inauspicious guidance.
Daily Fengshui Attire
A metaphysical dressing consultant that calculates the luckiest clothing colors based on the Chinese Earthly Branches and Five Elements of the day.
Bazi Pan - Eight Characters Fortune Telling
A sophisticated AI-driven charting tool for Bazi Eight Characters analysis using traditional Chinese metaphysics logic.
Batch Rename
A powerful script for batch renaming image datasets and their associated annotation files using flexible numbering and naming patterns.
Image Annotation QC
An automated quality control tool for validating, scoring, and visualizing image annotation accuracy in computer vision datasets.
Image Cropper
A high-performance utility for extracting specific objects from image datasets based on standard bounding box annotation formats.
Featured*
