Import & Export¶
The project's native format is .slp. Everything else is a conversion, handled
by sleap-io.js.
Opening projects¶
| Format | How |
|---|---|
.slp |
File ▸ Open Project…, or drag onto the window |
.pkg.slp |
Same — frames come out of the file, no video files needed |
Importing¶
File ▸ Import ▸ …
| Format | Notes |
|---|---|
| Analysis HDF5 | SLEAP's analysis .h5 export |
| NWB dataset | Neurodata Without Borders, ndx-pose |
| COCO dataset | COCO keypoint JSON |
| DeepLabCut dataset | A single DLC project |
| Multiple DeepLabCut datasets from folder | Batch-import a folder of DLC projects at once |
To bring in predictions produced outside the app, open their .slp with
File ▸ Open Project…, or merge it into the open project with
File ▸ Merge into Project….
Exporting¶
File ▸ Export ▸ …
| Format | What it's for |
|---|---|
| JSON | Plain-text dump of the labels |
| Analysis CSV | Tabular per-frame, per-node coordinates — the usual input to downstream analysis |
| Analysis HDF5 | The same data in SLEAP's analysis .h5 layout, for the current video only |
| NWB (ndx-pose) | Sharing and archiving in the NWB ecosystem. Desktop only |
| Labels Package | A .pkg.json file: the labels plus a list of the project's videos. No image data |
| Labeled Clip (Video) | An MP4 with the pose overlay rendered in |
For a portable project that carries its own frames, use the .pkg.slp labels
package below instead.
Labels packages¶
A labels package (.pkg.slp) embeds the image data alongside the labels, so
the project opens anywhere without its original videos. Export one with
Predict ▸ Export Labels Package… and pick a level:
| Level | Option | Contents |
|---|---|---|
| Level 1 | User labeled frames | Only frames you labeled by hand |
| Level 2 | User labeled + suggested frames | Your labeled frames plus suggested frames. The default |
| Level 3 | All labeled frames | Every labeled frame, including predictions |
Level 1 is the smallest package to send to a collaborator; Level 3 is a full archive.
Labeled clips¶
File ▸ Export ▸ Labeled Clip (Video)… renders MP4 with the skeleton overlay burned in. You pick the frame range, which videos to include, and the frame rate, and preview before rendering. Good for talks, figures, and showing someone what your data actually looks like.
Interoperability¶
.slp is the same format the legacy SLEAP GUI and the
Python sleap-io library read and write. You can label
here, analyze in Python, and open the same file in the Qt GUI, without converting
anything.
See the File Formats reference for the full table.