Skip to content

plugin-catalog.md

Latest commit

 

History

History
101 lines (74 loc) · 4.93 KB

plugin-catalog.md

File metadata and controls

101 lines (74 loc) · 4.93 KB

Plugin Catalog

IntegraPose ships with a curated plugin ecosystem that extends the core 7-tab workflow without bloating it. Plugins are opt-in — turn them on from Plugins → Manage Plugins..., launch them from the Plugins menu, and the rest of the app continues to work exactly as before.

The same project files, reproducibility bundles, and run manifests flow through plugins as through the main tabs, so a plugin-driven analysis stays as defensible as a core-only analysis.

!!! note "Plugin status — research in progress" The IntegraPose plugin ecosystem evolves alongside active research. Some plugins are stable enough for ongoing use; others are works in progress, and individual plugins may change, be deprecated, or be removed as research priorities shift. The set of bundled plugins reflects the current shipped state — not a long-term API contract. Pin to a commit hash if you depend on a specific plugin for an in-flight project.

Browse by purpose

Dataset creation

Plugin What it adds When to reach for it
Assisted Pose Curation Review-first pose labeling with model-assisted suggestions Building a pose dataset where each frame deserves a quick human pass
AutoLabel Forge GroundingDINO-driven autolabel + SAM-assisted cleanup for detection datasets Bootstrapping a detection dataset from raw video without box-by-box labelling
Dataset Augmentor Lab GUI-driven Albumentations augmentation for YOLO datasets Boosting data diversity before retraining when collection is the bottleneck

Behavior & sequence modeling

Plugin What it adds When to reach for it
BehaviorScope Toolkit LSTM-based behavior classifier with sequence prep, training, inference, and review Training a downstream behavior classifier from Tab 7's exported clip folders
Tandem YOLO Toolkit Coupled detector + temporal classifier orchestration Multi-stage classification projects where detection and behavior are decoupled

Domain-specific analytics

Plugin What it adds When to reach for it
Gait & Kinematic Dashboard Stride length, speed, paw angles, and group comparisons Locomotion-focused experiments and gait-disorder studies
Zone Counter Live polygon-based zone counts during inference Real-time entry/exit counts for arenas, choice tests, social zones

Exploration & review

Plugin What it adds When to reach for it
EDA Tool PCA / KMeans on pose embeddings, video-synced scatter plots Iterating on features and clusters with manuscript-ready exports

Browse by workflow stage

Workflow stage Useful plugins
Before annotation AutoLabel Forge, Assisted Pose Curation
Before training Dataset Augmentor Lab
After inference EDA Tool, Gait & Kinematic Dashboard, Zone Counter
After Tab 7 sub-behavior discovery BehaviorScope Toolkit, Tandem YOLO Toolkit

Browse by question

If you want to... Reach for
Speed up pose labeling Assisted Pose Curation
Build a detection dataset quickly AutoLabel Forge
Add augmentation diversity to a dataset Dataset Augmentor Lab
Inspect output tables visually EDA Tool
Count entries into a live region Zone Counter
Train a downstream behavior classifier BehaviorScope Toolkit
Run a multi-stage detector + classifier Tandem YOLO Toolkit
Focus on gait or stride features Gait & Kinematic Dashboard

Enabling and launching plugins

  1. Open Plugins → Manage Plugins... to see installed plugins.
  2. Tick the plugins you want, accept the trust dialog if shown, and apply.
  3. Each enabled plugin appears in the Plugins menu and launches in its own window.

Most plugins write into the same project / output folder you already use, so there is no extra bookkeeping.

Creating your own plugins

Labs can add local workflow extensions without editing the core IntegraPose tabs. See Plugin Development for the plugin folder layout, plugin.json metadata, plugin.py registration entry points, opt-in/trust behavior, and a minimal working template.

Practical notes

  • The main IntegraPose workflow remains:
Data Preprocessing -> Setup -> Training -> Inference -> Bout Analytics -> Behavior Clustering

Plugins are accelerators or specialty tools, not required steps for every project.

  • A few plugins have additional dependencies (for example, AutoLabel Forge wants GroundingDINO and SAM weights, and BehaviorScope wants a recent PyTorch). Per-plugin pages call those out where they apply.

  • All plugins respect the IntegraPose seed configuration and project-level settings, so reproducibility bundles stay consistent across plugin and core runs.