Using the GUI¶

The SLEAP labeling interface is launched via the sleap label command (or legacy sleap-label). See Command Line Interfaces for details.

Keyboard shortcuts

Most menu commands have keyboard shortcuts. View and customize them via Help → Keyboard Shortcuts.

Menus¶

File¶

Command	Description
New...	Create a new project
Open...	Open an existing project
Save / Save As...	Save the current project
Import...	Import from external formats: COCO (`.json`), DeepLabCut (`.csv`), DeepPoseKit (`.h5`), LEAP (`.mat`)
Merge Data From...	Copy labels/predictions from another SLEAP project into the current one
Add Videos...	Add videos to the current project
Replace Videos...	Swap videos with copies at different paths (useful for moving between drives)

Go¶

Command	Description
Next Labeled Frame	Jump to next frame with any labels (user or predicted)
Next User Labeled Frame	Jump to next frame with user labels only
Previous User Labeled Frame	Jump to previous frame with user labels only
Next Suggestion	Jump to the next suggested frame
Next Track Spawn Frame	Jump to where a new track starts (useful for proofreading)
Next Video	Switch to the next video in the project
Go to Frame...	Jump to a specific frame number
Select to Frame...	Select a clip from current frame to a specified frame

View¶

Command	Description
Color Predicted Instances	Toggle coloring predictions by track (enable for proofreading)
Color Palette	Choose colors for instances (see note below)
Apply Distinct Colors To	Color by tracks, nodes, or edges
Show Instances	Toggle visibility of all instances
Show Non-Visible Nodes	Toggle visibility of occluded/missing nodes
Show Node Names	Toggle node name labels
Show Edges	Toggle edge visibility
Edge Style	Lines or wedges (wedges show orientation)
Trail Length	Show instance movement trails (useful for detecting swaps)
Fit Instances to View	Auto-zoom to instances in each frame
Seekbar Header	Plot information above the seekbar
Crop Size Overlay	Show the crop region for top-down training pipelines

Color palettes

"alphabet" has 26 visually distinct colors
Palettes ending with "+" don't cycle (useful for proofreading—"five+" shows track 4+ as orange)
Customize palettes in ~/.sleap/colors.yaml

Labels¶

Command	Description
Add Instance	Add an instance to the current frame
Instance Placement Method	Choose how new instances are positioned ("Best" uses predictions first)
Delete Instance	Delete the selected instance
Set Instance Track	Assign selected instance to a different track
Propagate Track Labels	Apply track changes to all subsequent frames
Transpose Instance Tracks	Swap tracks between two instances
Delete Instance and Track	Delete all instances in a track across all frames
Custom Instance Delete...	Delete instances matching specific criteria
Select Next Instance	Cycle through instances in the frame
Clear Selection	Deselect the current instance

Predict¶

Command	Description
Run Training...	Train models from your labeled data
Run Inference...	Generate predictions using trained models
Evaluate Metrics for Trained Models...	View recall, precision, and other metrics
Add Instances from All Predictions on Current Frame	Convert all predictions to editable instances
Delete All Predictions...	Remove all predictions from current video
Delete All Predictions from Clip...	Remove predictions from selected frame range
Delete All Predictions from Area...	Remove predictions within a rectangular region
Delete All Predictions with Low Score...	Remove predictions below a score threshold
Delete All Predictions beyond Frame Limit...	Keep only top N predictions per frame
Delete Predictions on User-Labeled Frames	Remove predictions from frames with user labels
Export Video with Visual Annotations...	Export video with poses overlaid

Random sample (current video)

When running inference, you can now choose "Random sample (current video)" to predict on a random subset of frames from just the current video.

Analyze¶

Command	Description
Instance Size Distribution...	View bounding box size distribution (helps choose crop sizes)
Label QC...	Open the Label Quality Control panel to find annotation errors

Help¶

Command	Description
Keyboard Shortcuts	View and customize shortcuts
Check for Updates	Check GitHub for newer SLEAP versions and shows a dialog with release notes if an update is available

Analysis Widgets¶

Instance Size Distribution¶

Access via Analyze → Instance Size Distribution... or the toolbar.

This widget shows the distribution of instance bounding box sizes in your labeled data. Use it to:

Choose crop sizes for top-down models—crop size should encompass most instances
Identify outliers that may indicate labeling errors
Verify consistency across your labeled frames

For a detailed walkthrough including rotation augmentation, statistics interpretation, and programmatic access, see Instance Size Distribution Guide.

Crop Size Overlay¶

Enable via View → Crop Size Overlay.

Shows the crop region that will be used for top-down training pipelines, similar to the receptive field overlay. Helps verify your crop size setting captures the full animal.

Label QC¶

Access via Analyze → Label QC...

This panel uses statistical anomaly detection to automatically find potential labeling errors in your dataset. It flags instances with unusual edge lengths, joint angles, node spacing, or other features that deviate from the norm.

Use it to:

Catch errors before training—find labeling mistakes early
Verify proofreading—check tracking corrections after proofreading
Identify systematic issues—see patterns in labeling errors

For a complete guide including sensitivity tuning, issue types, and programmatic access, see Label Quality Control Guide.

Mouse Controls¶

Action	How
Zoom in/out	Mouse wheel on image
Pan image	Left-click + drag
Toggle node visibility	Right-click on node
Add instance	Right-click elsewhere on image
Zoom to region	Alt + left-click + drag
Zoom out	Alt + double-click
Move entire instance	Alt + drag on node
Rotate entire instance	Alt + mouse wheel on node
Create instance from prediction	Double-click on predicted instance
Add missing nodes to instance	Double-click on editable instance
Select instance	Click on instance
Clear selection	Click elsewhere
Duplicate instance	Ctrl + drag on instance

macOS

Substitute Option for Alt and Command for Control.

Key	Action
→ / ←	Move one frame forward/back
Ctrl + → / ←	Move medium step (4 frames)
Ctrl + Alt + → / ←	Move large step (100 frames)
Home / End	First / last frame
Shift + navigation	Select frames while moving
1-9	Select instance by number
Ctrl (hold)	Show tracks legend
Escape	Deselect all

Seekbar Controls¶

Action	How
Select frame range	Shift + drag
Clear selection	Shift + click
Zoom to range	Alt + drag
Zoom out (show all)	Alt + click

Labeling Suggestions¶

Generate suggested frames for labeling via Labeling Suggestions panel:

Method	Description
Sample	Evenly spaced ("stride") or random frames from each video
Image Features	Visually distinctive frames for training diversity (slower)
Prediction Score	Frames with low-confidence predictions (for proofreading)
Velocity	Frames where instances move unusually fast (may indicate tracking errors)