Get Started

The recommended way to run CartLoader is run_together: a single command that takes a sample from raw platform output to deployable CartoScope assets: SGE conversion, FICTURE spatial factors, cell analyses, background images, and PMTiles packaging. It plans every step as one resumable Makefile, so rerunning the same command after an interruption picks up where it stopped.

Both quick starts below run two examples:

1. A small mouse hippocampus dataset (Seq-Scope), finishes in minutes. In Docker it needs no download at all.
2. A Xenium human kidney section (GSE264334), which exercises the whole pipeline, including cell segmentation imports and the DAPI image.

Choose how to run

• • 🚀 Run with Docker

    Run cartloader inside the prebuilt image hyunminkang/cartloader.

    • ✅ No installation required
    • 📦 All tools and dependencies included
    • 🧪 Starter dataset preloaded

  Read

• • 🔧 Run Locally

    Install cartloader and its tools on your system.

    • 💻 Use the latest code from GitHub
    • ⚙️ Full flexibility
    • 📂 Easy to customize and develop

  Read

Not sure which to choose?

See Docker or local?. In short: use Docker to try CartLoader or to run a released version reproducibly; install locally to follow the latest code or modify it.

After the quick start

• Your own data: the platform tutorials show the run_together command for each supported platform, and Specifying Inputs covers every way to describe a sample.
• Several samples together: Multiple Samples with run_together trains one shared model across samples.
• Publishing and annotation: run_together --s3-upload and --anno / --anno-deep (see Publishing).
• Individual commands: the Step-by-step section runs each stage (sge_convert, run_ficture2, run_cartload2, ...) separately for full control.