Skip to content

Visualization (CryoSiam‑Vis)

The cryosiam-vis package provides interactive visualization utilities (via napari and dash) for results produced by CryoSiam:

  • Denoised tomograms
  • Semantic segmentation results
  • Instance segmentation results
  • Particle/coordinate sets from .star files
  • Embedding UMAP spaces

Install

Refer to the CryoSiam-Vis installation guide.

Command Overview

All visualizations are accessed through a single CLI with subcommands:

cryosiam-vis --version
cryosiam-vis <subcommand> [arguments]

Available subcommands:

  • visualize_denoising — show denoised vs raw tomogram
  • visualize_semantic — overlay semantic masks on tomogram
  • visualize_instance — explore instance masks
  • visualize_coordinates — plot points from a .star file on a tomogram
  • visualize_embeddings — open an interactive embedding explorer

visualize_denoising

Show the denoised tomogram produced by cryosiam denoise (and optionally compare to raw).

Usage

cryosiam-vis visualize_denoising --config_file configs/denoise.yaml --filename TS_01.mrc

Arguments

Arg Required Description
--config_file ✅ The same YAML used with cryosiam denoise_predict (paths are read from it).
--filename ✅ Tomogram filename (including extension), must exist under data_folder from the config.

What it loads

  • Input tomogram from data_folder
  • Denoised output from prediction_folder
  • Opens layers in napari for side‑by‑side inspection

visualize_semantic

Overlay semantic predictions over the denoised tomogram.

Usage

cryosiam-vis visualize_semantic --config_file configs/semantic.yaml --filename TS_01.mrc

Arguments

Arg Required Description
--config_file ✅ The same YAML used with cryosiam semantic_predict. Reads data_folder, mask_folder, and prediction_folder.
--filename ✅ Tomogram filename (including extension).

What it loads

  • Denoised tomogram (data_folder)
  • Semantic probabilities and/or segmentation (prediction_folder)

visualize_instance

Visualize instance segmentation volumes and labels.

Usage

cryosiam-vis visualize_instance --config_file configs/instance.yaml --filename TS_01.mrc

Arguments

Arg Required Description
--config_file ✅ The same YAML used with cryosiam instance_predict.
--filename ✅ Tomogram filename (including extension).

What it loads

  • Denoised tomogram (data_folder)
  • Instance masks/labels from the prediction folder defined in the config

visualize_coordinates

Plot particle coordinates from a .star file onto a tomogram.

Usage

cryosiam-vis visualize_coordinates --tomo_path /path/to/denoised --filename TS_01.mrc --star_file /path/to/coordinates.star --point_size 15

Arguments

Arg Required Default Description
--tomo_path ✅ — Folder containing tomograms (ideally denoised).
--filename ✅ — Tomogram filename (with extension).
--star_file ✅ — Path to the .star file with point coordinates.
--point_size ❌ 15 Point marker size in napari.

Notes

  • Make sure the voxel size / coordinate system used to generate the .star matches the tomogram voxel size.
  • Supports common STAR conventions via starfile.

visualize_embeddings

Open the embedding viewer (for exploratory analysis).

Usage

cryosiam-vis visualize_embeddings

What it does

  • Loads embedding UMAP (if present in the working directory or a default location used by your inference)
  • Opens an interactive explorer (2D UMAP projection, selection, metadata overlays)

See also