Skip to content

Where things live

MindSight keeps files in two distinct places, and knowing which is which saves a lot of confusion. This page is the map.

Two locations, don't mix them up

  • The install root (~/MindSight on macOS, %LOCALAPPDATA%\MindSight on Windows) holds the application itself -- its private Python, the model weights, and the default data folders.
  • The app-state directory (~/.mindsight, on both platforms) holds your personal app settings -- presets, last-used session, recent projects.

They are separate on purpose: reinstalling or moving the app touches the install root; it does not disturb your settings in ~/.mindsight.

The install root

The installer puts everything under one folder:

  • macOS -- ~/MindSight (your home folder's MindSight directory).
  • Windows -- %LOCALAPPDATA%\MindSight (usually C:\Users\<you>\AppData\Local\MindSight).
MindSight/                 # the install root
├── venv/                  # a private Python + all of MindSight's dependencies
└── app/                   # the application data home
    ├── Weights/           # model weights, one subfolder per backend (Weights/YOLO/, Weights/MGaze/, ...)
    ├── Outputs/           # default output location for runs that don't target a project
    ├── Projects/          # the sample ExampleStudy and any projects you keep here
    └── ...                # the app source and assets

Weight resolution is global: a bare filename like yolov8n.pt resolves to Weights/<backend>/yolov8n.pt for every project on the machine, which is why the Models tab and preflight talk about one shared Weights/ folder rather than a per-project one.

The app-state directory

Your personal settings live in ~/.mindsight (a hidden folder in your home directory, on both macOS and Windows):

Path What it holds
~/.mindsight/presets/ Named setting bundles saved via File > Save Preset....
~/.mindsight/last_used.json The last session's settings, restored on the next launch.
~/.mindsight/recent_projects.json The recent-projects list shown on the Projects tab.
~/.mindsight/run_settings.json The persisted Inference Settings store (the dialog's state).

Because this is separate from the install root, your presets and recent-projects list survive an app reinstall.

A project folder

A study project is a self-contained folder (see Projects and sessions). The sample Projects/ExampleStudy/ shows the shape:

MyProject/
├── project.yaml            # points at the pipeline preset; holds participants/conditions
├── notes.md                # free-form study notes (shown on the Projects tab)
├── Pipeline/
│   └── pipeline.yaml       # the pipeline preset this study runs with
├── Inputs/
│   ├── Runs/               # one subfolder per session: Runs/<run_id>/ with one video + run.yaml
│   │                       #   (a Runs/<run_id>/ with run.yaml but NO video is a planned session)
│   ├── Videos/             # (legacy flat layout) recordings directly in a folder
│   └── Prompts/            # the study's .vp.json visual prompt, if used
└── Outputs/                # created when you Run
    ├── Runs/<run_id>/      # per-run CSVs and outputs (run-folder projects)
    ├── CSV Files/          # per-video CSVs + the project-wide Global_* aggregates (flat projects)
    ├── Videos/             # annotated output videos
    ├── heatmaps/           # heatmap images
    └── _run/
        └── ledger.json     # the resume ledger: what's done, and archived prior outputs

A project uses one input layout: run-folder (Inputs/Runs/<run_id>/, what the wizard builds and the recommended layout) or the legacy flat layout (Inputs/Videos/). Do not mix both -- preflight stops you if a project has videos in both places.

The _run/ledger.json file is what makes runs resumable: it records, per recording, whether it is done and unchanged, and it archives superseded outputs rather than deleting them.

Quick-analysis outputs

Runs done without a project -- Video File mode, or a Camera quick-run -- do not use the project tree. They write to a folder you choose (prefilled next to the source in Video File mode, and editable). A Camera quick-run also drops a <run_id>_session.yaml sidecar next to its outputs, carrying the session's metadata so the recording can be imported into a project later.

See also