see-spot

Overview

seespot-demo.mp4

Installation

CodeOcean

Clone this Capsule

Follow instructions

Local install

# Clone repository
git clone https://github.com/AllenNeuralDynamics/see-spot.git
cd see-spot

# Run installer (uses defaults: port 5555, cache at ~/.seespot/cache)
./install.sh

# Or customize settings interactively
./install.sh --interactive

# Start the server
seespot start

# Access at http://localhost:5555

For detailed installation instructions, AWS credentials setup, and troubleshooting, see INSTALL.md.

Development Setup

• git clone https://github.com/AllenNeuralDynamics/see-spot.git
• Install: uv sync
• Launch with auto-reload:

cd /home/matt.davis/code/see-spot && source .venv/bin/activate && cd src && uvicorn see_spot.app:app --host 0.0.0.0 --port 9999 --reload

Dataset Info

Regular (Fused) Datasets

Standard datasets with fused image data at the top level:

dataset_name/
  image_spot_spectral_unmixing/
    mixed_spots_*.pkl
    unmixed_spots_*.pkl
  image_tile_fusing/fused/
    channel_*.zarr

Tiled (Non-Fused) Datasets

Datasets with independent tile processing, where each tile has separate spot data:

dataset_name/
  image_spot_spectral_unmixing/
    Tile_X_0001_Y_0000_Z_0000/
      mixed_spots_*_tile_*.pkl
      unmixed_spots_*_tile_*.pkl
    Tile_X_0002_Y_0000_Z_0000/
      ...

When downloading a tiled dataset, the system automatically:

• Detects tile subfolders (beginning with "Tile")
• Creates virtual dataset entries for each tile
• Names them as: {dataset_name}_X_####_Y_####_Z_####

Changelog

• v0.7.0 (11-21-2025)
  • Big feature: loading of single tile dataset (non-fused)
    • Looks in spectral_unmixing, if "Tile" subfolders" load pkl files from each and show them in the app
    • Creates virtual datasets for each tile, shown in dataset management table
  • Big feature: use lasso to make multi selection neuroglancer annotations
  • Legend series visibilty persists across app interactions
  • class balanced sampling option
  • removed spots in unmixed hidden by default
  • load mixed table first
  • dye lines plotting
  • display filters (r dist) refactor
  • Better dataset management (with filters) use DataTables js
  • add spot persistent marker when clicked
• v0.5.0 (09-19-2025)
  • backend downloads mixed + unmixed tables, merges and saves as .parquet (massive compression)
  • polars for dataframe manipulation (huge speedup)
  • optimized data types in tables, reduced memory demands
  • added removed_spots key in data table and plotting in frontend
  • Added mixed/unmixed channel display mode toggle with dedicated UI controls
  • Implemented chart axis limit controls (auto, fixed, min/max, percentile modes)
  • Added Sankey flow diagram showing channel reassignment patterns with backend data calculation
  • Enhanced data filtering with valid spots toggle and removed spots highlighting

Contributing

Linters and testing

coverage run -m unittest discover && coverage report
interrogate .
flake8 .
black .
isort .

Pull requests

• Internal members, please create a branch.
• External members, fork repo and open PR

Commit style

• We primarily use Angular style for commit messages. Roughly, they should follow the pattern:
• <type>: <short summary>

Type:

• build: Changes that affect build tools or external dependencies (example scopes: pyproject.toml, setup.py)
• ci: Changes to our CI configuration files and scripts (examples: .github/workflows/ci.yml)
• docs: Documentation only changes
• feat: A new feature
• fix: A bugfix
• perf: A code change that improves performance
• refactor: A code change that neither fixes a bug nor adds a feature
• test: Adding missing tests or correcting existing tests