Skip to content

📌 Welcome to the Picsellia CV Engine Docs

Picsellia CV Engine is a modular toolkit for building, testing, and deploying computer vision pipelines — all fully integrated with Picsellia. Whether you're processing datasets, training models, or deploying experiments, this engine helps you structure everything in a clean, reusable way.

🧠 What’s a pipeline?

A pipeline is a sequence of steps that defines how data flows — from raw inputs to final results.

In Picsellia CV Engine, pipelines are used for both:

  • Training pipelines:

    Load training datasets, configure a model, run training, log results and export weights.

  • Processing pipelines:

    Clean or filter datasets, apply data augmentation, run inference for pre-annotation, or convert formats.

Each unit of work is a step — a standalone function decorated with @step. You can reuse, extend, or combine steps freely depending on your needs.

✨ Key features

  • Composable Steps – Use or customize ready-made steps for common operations (loading data, training, etc.)
  • Training Pipelines – Structure model training (e.g. Ultralytics YOLO) with built-in logic
  • Processing Pipelines – Clean, transform, or validate datasets before use
  • Framework Extensions – Support custom training libraries via a pluggable architecture
  • CLI Automation – Use pxl-pipeline cli to scaffold, test, and deploy pipelines locally or on Picsellia

🚀 Get started

  • 📦 Installation Guide – Set up the engine and CLI
  • 🛠 Usage Guide – Build your first processing or training pipeline
  • 📖 API Reference – Explore contexts, decorators, steps, and framework integrations

👋 New to Picsellia?