What happens to my data if I stop the container?

Your datasets are preserved in the MongoDB sidecar and the configuration directory, ensuring persistence across restarts.

Can I use my GPU for model inference in FiftyOne?

Yes, the skill supports GPU acceleration and can auto-detect or be explicitly configured for NVIDIA, AMD, and Intel GPUs.

Does this skill require a separate database installation?

No, the FiftyOne skill automatically manages a MongoDB sidecar container to handle persistent dataset storage and metadata.

How do I access the FiftyOne web interface?

You can find the access link by running 'ujust fiftyone url', which will provide the local or network address for the web UI.

How do I mount my local datasets into the FiftyOne container?

Use the 'ujust fiftyone config --workspace-dir=/path/to/data' command to map your local directory to the container's workspace.

FiftyOne ML Dataset Curator

Name: FiftyOne ML Dataset Curator
Author: atrawog

byatrawog

Data Science & ML

Manages FiftyOne dataset visualization and curation workflows using Podman Quadlet containers with MongoDB persistence.

About

This skill provides a comprehensive management interface for FiftyOne, the leading open-source tool for building high-quality datasets and computer vision models. It automates the deployment of a multi-container architecture, including a MongoDB sidecar for persistent storage and full GPU acceleration support on Bazzite AI. By leveraging Podman Quadlets, it allows users to easily configure, start, and scale dataset analysis environments, manage plugins, and interact with datasets via a web UI or direct container shell, making it an essential tool for machine learning engineers and data scientists.

Key Features

Automatic GPU acceleration support for NVIDIA, AMD, and Intel hardware
Persistent dataset storage and metadata management
Multi-container orchestration with FiftyOne and MongoDB sidecar
Integrated plugin management for extending FiftyOne functionality
Flexible workspace mounting and network binding configurations
0 GitHub stars

Use Cases

Analyzing model inference results and visualizing detection errors
Curating and labeling training data for machine learning pipelines
Visualizing and exploring large-scale computer vision datasets

About

Key Features

Automatic GPU acceleration support for NVIDIA, AMD, and Intel hardware
Persistent dataset storage and metadata management
Multi-container orchestration with FiftyOne and MongoDB sidecar
Integrated plugin management for extending FiftyOne functionality
Flexible workspace mounting and network binding configurations
0 GitHub stars

Use Cases

Analyzing model inference results and visualizing detection errors
Curating and labeling training data for machine learning pipelines
Visualizing and exploring large-scale computer vision datasets