Note
Go to the end to download the full example code.
Convert Lightning Pose project to the PoseMark format#
Prepare a Lightning Pose (LP) project to be contributed to the PoseMark corpus.
Imports#
import json
import shutil
import tempfile
from datetime import datetime, timezone
from pathlib import Path
import poseinterface
from poseinterface.clips import extract_single_clip
from poseinterface.io import (
annotations_to_posemark,
frames_to_posemark,
predictions_to_posemark,
split_lp_collected_data,
video_to_posemark,
)
from poseinterface.utils import tree
Overview#
We’ll handle the conversion in two steps:
Convert: LP project files (videos, frame annotations, and keypoint predictions) are restructured into the PoseMark layout.
Extract clips: Short video clips and their labels are extracted from the converted videos and their corresponding keypoint predictions, ready for expert review.
The workflow is similar to the one followed in Convert DeepLabCut project to the PoseMark format, with a few differences explained below.
Source Lightning Pose project#
We work with a dataset from the International Brain Laboratory (IBL), containing videos of mouse paw movements analysed using Lightning Pose.
Note
This example runs against a lightweight fixture shipped with the
repository (under tests/data/). Replace source_project_dir
and posemark_base_dir with the paths to your LP project and
PoseMark directories, respectively. Keep in mind that your
project will contain more files than are shown here.
Warning
Lightning Pose saves prediction files to the model output directory,
not to the project’s videos/ directory. Before running this
script, move (or copy) each session’s prediction CSV — and apply any
manual corrections you have made — into
<source_project_dir>/videos/, named to match the corresponding
video stem (e.g. <video_stem>.csv).
source_project_dir = (
Path(".").resolve().parent / "tests" / "data" / "lightningpose" / "ibl-paw"
)
print(tree(source_project_dir, level=1, exclude_hidden=True))
# For this example we use a temporary directory, cleaned up at the end.
posemark_base_dir = Path(tempfile.mkdtemp(prefix="posemark-"))
print(f"\nPoseMark dataset will be saved to: {posemark_base_dir}")
ibl-paw/
├── CollectedData.csv
├── labeled-data/
└── videos/
2 directories, 1 files
PoseMark dataset will be saved to: /tmp/posemark-76fx4td3
The LP project differs from a DLC project in one key respect: all session
annotations live in a single project-level CollectedData.csv rather
than in per-session files inside labeled-data/. The two sub-directories
we care about otherwise mirror the DLC layout:
videos/: session videos and (after the move described above) their corresponding prediction CSVs.labeled-data/: sampled frames.
print(tree(source_project_dir / "videos", level=1, exclude_hidden=True))
videos/
├── 6c6983ef73834989918332b1a300d17a_left.csv
├── 6c6983ef73834989918332b1a300d17a_left.mp4
├── a92c4b1d46bd457ea1f4414265f0e2d4_left.csv
└── a92c4b1d46bd457ea1f4414265f0e2d4_left.mp4
0 directories, 4 files
print(tree(source_project_dir / "labeled-data", level=2, exclude_hidden=True))
labeled-data/
├── 6c6983ef73834989918332b1a300d17a_left/
│ ├── img00000006.png
│ ├── img00000036.png
│ └── img00000136.png
└── a92c4b1d46bd457ea1f4414265f0e2d4_left/
├── img00014918.png
├── img00074921.png
└── img00074925.png
2 directories, 6 files
Define sessions to convert#
We select two sessions from the LP project and assign each to either
the Train or Test split of the
PoseMark.
You may expand this list with more sessions, but ensure that each session
belongs to exactly one split, and that the same subject doesn’t appear in
both splits (to avoid data leakage).
sessions = [
{
"split": "Train",
"source_video": "6c6983ef73834989918332b1a300d17a_left.mp4",
"sub_id": "SWC054",
"ses_id": "6c6983ef73834989918332b1a300d17a",
"cam_id": "left",
},
{
"split": "Test",
"source_video": "a92c4b1d46bd457ea1f4414265f0e2d4_left.mp4",
"sub_id": "KS023",
"ses_id": "a92c4b1d46bd457ea1f4414265f0e2d4",
"cam_id": "left",
},
]
project_name = "IBL-paw"
Split the project-level annotation file#
Unlike DLC, Lightning Pose stores all session annotations in a single
project-level CollectedData.csv. We split it into per-session
CollectedData_<scorer>.csv files and create a temporary directory
mirroring the labeled-data/ structure with copies of the original
frames. This is necessary because the underlying loader (sleap-io)
resolves image paths relative to the CSV location, so the split CSV
must live alongside the frame images it references.
lp_session_base = posemark_base_dir / ".lp_sessions"
split_results = split_lp_collected_data(
input_path=source_project_dir / "CollectedData.csv",
output_dir=lp_session_base,
)
for ses_name, split_csv in split_results.items():
orig_frames_dir = source_project_dir / "labeled-data" / ses_name
ses_dir = split_csv.parent
for img in sorted(orig_frames_dir.glob("*.png")):
shutil.copy(img, ses_dir / img.name)
print("Split annotation files:")
for ses_name, csv_path in split_results.items():
print(f" {ses_name}: {csv_path.name}")
Split annotation files:
6c6983ef73834989918332b1a300d17a_left: CollectedData_mattw.csv
a92c4b1d46bd457ea1f4414265f0e2d4_left: CollectedData_mattw.csv
Convert to PoseMark format#
For each session we:
copy (and re-encode, if necessary) the session video;
convert LP keypoint annotations to COCO JSON, as well as copy and rename the corresponding frame images;
convert LP keypoint predictions to COCO JSON.
for session in sessions:
split = session["split"]
ids = {k: session[k] for k in ["sub_id", "ses_id", "cam_id"]}
sub_ses_prefix = f"sub-{ids['sub_id']}_ses-{ids['ses_id']}"
sub_ses_cam_prefix = f"{sub_ses_prefix}_cam-{ids['cam_id']}"
source_video_path = source_project_dir / "videos" / session["source_video"]
target_session_dir = (
posemark_base_dir / split / project_name / sub_ses_prefix
)
target_frames_dir = target_session_dir / "Frames"
target_frames_dir.mkdir(parents=True, exist_ok=True)
# LP session name matches ses_id + "_" + cam_id (e.g. labeled-data/ dir).
video_stem = source_video_path.stem
_lp_key = f"{session['ses_id']}_{session['cam_id']}"
lp_session_name: str | None = _lp_key if _lp_key in split_results else None
print(f"Converting session: {split}/{project_name}/{sub_ses_prefix}")
# Copy the session video, re-encoding to H.264/yuv420p if necessary.
video_to_posemark(
input_video=source_video_path,
output_video_dir=target_session_dir,
**ids,
)
print(f"\tvideo: {source_video_path.name} -> {sub_ses_cam_prefix}.mp4")
# Convert LP annotations to COCO frame labels JSON, then copy the
# corresponding frame images with standardised PoseMark filenames.
if lp_session_name is None:
print(
f"\tNo matching LP session found for {video_stem!r}."
" Skipping annotations-to-PoseMark conversion."
)
else:
# The split CSV lives in the temp dir alongside copied frames so
# that sleap-io can resolve image paths relative to the CSV location.
source_annotations_path = split_results[lp_session_name]
# Use the original frames dir (not the copies) for the copy step.
source_frames_dir = (
source_project_dir / "labeled-data" / lp_session_name
)
framelabels_path = annotations_to_posemark(
input_path=source_annotations_path,
output_dir=target_frames_dir,
format="frame",
**ids,
)
frames_to_posemark(
input_dir=source_frames_dir,
output_dir=target_frames_dir,
framelabels_path=framelabels_path,
)
print(
f"\tannotations (+ frame images): "
f"{source_annotations_path.name} -> {framelabels_path.name}"
)
# Convert LP predictions to COCO video labels JSON for clip extraction.
# Prediction CSVs must be present in videos/ before running this script;
# see the warning in the "Source Lightning Pose project" section above.
source_predictions_path = next(
(source_project_dir / "videos").glob(f"{video_stem}*.csv"),
None,
)
if source_predictions_path is None:
print(
f"\tNo prediction CSV found for {video_stem!r} in "
f"{source_project_dir / 'videos'}. Skipping predictions-to-"
"PoseMark conversion."
)
else:
predictions_to_posemark(
input_path=source_predictions_path,
video_path=source_video_path,
output_dir=target_session_dir,
**ids,
)
print(
f"\tpredictions: {source_predictions_path.name} -> "
f"{sub_ses_cam_prefix}_videolabels.json"
)
print("Done.\n")
Converting session: Train/IBL-paw/sub-SWC054_ses-6c6983ef73834989918332b1a300d17a
video: 6c6983ef73834989918332b1a300d17a_left.mp4 -> sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left.mp4
annotations (+ frame images): CollectedData_mattw.csv -> sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_framelabels.json
predictions: 6c6983ef73834989918332b1a300d17a_left.csv -> sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_videolabels.json
Done.
Converting session: Test/IBL-paw/sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4
video: a92c4b1d46bd457ea1f4414265f0e2d4_left.mp4 -> sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left.mp4
annotations (+ frame images): CollectedData_mattw.csv -> sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_framelabels.json
predictions: a92c4b1d46bd457ea1f4414265f0e2d4_left.csv -> sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_videolabels.json
Done.
The resulting PoseMark dataset:
print(tree(posemark_base_dir, level=5, exclude_hidden=True))
posemark-76fx4td3/
├── Test/
│ └── IBL-paw/
│ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4/
│ ├── Frames/
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-14918.png
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-74921.png
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-74925.png
│ │ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_framelabels.json
│ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left.mp4
│ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_videolabels.json
└── Train/
└── IBL-paw/
└── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a/
├── Frames/
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-006.png
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-036.png
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-136.png
│ └── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_framelabels.json
├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left.mp4
└── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_videolabels.json
8 directories, 12 files
Note
Frame labels (framelabels.json) are generated for both splits,
but in the published dataset the Test split intentionally
omits them for evaluation. See the folder structure
specification for details.
The videolabels.json files generated alongside each session video
are intermediate artifacts used for clip extraction in the next
section, and will not be included in the published dataset.
Extract clips#
Clips (short video segments) can be extracted from the converted session
videos. When the videolabels.json files are present, the corresponding
clip label files (cliplabels.json) are generated automatically during
clip extraction.
These clip label files should then be proof-read and corrected by
experts before being included in PoseMark.
First, we specify the clip-extraction parameters. This step can be repeated with different parameters to incrementally expand the clip set.
duration = 5 # in frames
start_frames = [25, 50, 75]
print(f"Extracting {duration}-frame clips starting at frames: {start_frames}")
Extracting 5-frame clips starting at frames: [25, 50, 75]
We loop over all sessions and extract clips at each start frame.
The resulting video clips and their cliplabels.json files are saved
in a Clips/ subdirectory within each session folder.
for session in sessions:
sub_ses_prefix = f"sub-{session['sub_id']}_ses-{session['ses_id']}"
sub_ses_cam_prefix = f"{sub_ses_prefix}_cam-{session['cam_id']}"
session_dir = (
posemark_base_dir / session["split"] / project_name / sub_ses_prefix
)
for start_frame in start_frames:
clip_path, _ = extract_single_clip(
video_path=session_dir / f"{sub_ses_cam_prefix}.mp4",
start_frame=start_frame,
duration=duration,
)
print(f"Extracted clip: {clip_path.stem}")
Extracted clip: sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-25_dur-5
Extracted clip: sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-50_dur-5
Extracted clip: sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-75_dur-5
Extracted clip: sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-25_dur-5
Extracted clip: sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-50_dur-5
Extracted clip: sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-75_dur-5
The resulting PoseMark dataset, including the extracted clips and their corresponding labels:
print(tree(posemark_base_dir, level=5, exclude_hidden=True))
posemark-76fx4td3/
├── Test/
│ └── IBL-paw/
│ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4/
│ ├── Clips/
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-25_dur-5.mp4
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-25_dur-5_cliplabels.json
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-50_dur-5.mp4
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-50_dur-5_cliplabels.json
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-75_dur-5.mp4
│ │ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_start-75_dur-5_cliplabels.json
│ ├── Frames/
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-14918.png
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-74921.png
│ │ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_frame-74925.png
│ │ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_framelabels.json
│ ├── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left.mp4
│ └── sub-KS023_ses-a92c4b1d46bd457ea1f4414265f0e2d4_cam-left_videolabels.json
└── Train/
└── IBL-paw/
└── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a/
├── Clips/
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-25_dur-5.mp4
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-25_dur-5_cliplabels.json
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-50_dur-5.mp4
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-50_dur-5_cliplabels.json
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-75_dur-5.mp4
│ └── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_start-75_dur-5_cliplabels.json
├── Frames/
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-006.png
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-036.png
│ ├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_frame-136.png
│ └── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_framelabels.json
├── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left.mp4
└── sub-SWC054_ses-6c6983ef73834989918332b1a300d17a_cam-left_videolabels.json
10 directories, 24 files
Note
In the published dataset, the Train split includes all
cliplabels.json files, while the Test split omits them. Clip
start labels (startlabels.json) are automatically derived during
data submission from each clip’s first frame, and are included in
both splits to support point-tracker evaluation.
The videolabels.json files generated in the previous section are
intermediate artifacts used for clip extraction, and are never shared.
See the folder structure specification for details.
Record provenance (optional)#
This step is optional and can be safely skipped, but it is highly recommended when converting real data, for book-keeping and reproducibility purposes.
We save a copy of this script alongside a JSON sidecar with the
poseinterface version (including git commit, via setuptools_scm)
and a UTC timestamp. Both files are written to a top-level
.provenance/ folder, named by project, so multiple projects under
the same posemark_base_dir stay distinct.
# sphinx_gallery_capture_repr = ()
provenance_dir = posemark_base_dir / ".provenance"
provenance_dir.mkdir(parents=True, exist_ok=True)
# ``__file__`` is set when running this script directly with Python, but
# not when sphinx-gallery executes it during the docs build.
script_path_str = globals().get("__file__")
if script_path_str:
shutil.copy(Path(script_path_str), provenance_dir / f"{project_name}.py")
(provenance_dir / f"{project_name}.json").write_text(
json.dumps(
{
"poseinterface_version": poseinterface.__version__,
"converted_at": datetime.now(timezone.utc).isoformat(),
"source_project_dir": str(source_project_dir),
},
indent=2,
)
)
Clean up#
Since this example writes to a temporary directory, we remove it at the end.
Warning
Only run this cell when posemark_base_dir points to a temporary
location. The guard below refuses to delete anything outside the
system temp directory, so it is safe to leave in place when you adapt
this example to a real PoseMark dataset path.
system_tempdir = Path(tempfile.gettempdir()).resolve()
target = posemark_base_dir.resolve()
if target.is_relative_to(system_tempdir) and target != system_tempdir:
shutil.rmtree(target)
print(f"Removed temporary PoseMark directory: {target}")
else:
print(
f"Refusing to remove {target}: not inside system temp dir "
f"({system_tempdir}). Delete manually if you really want to."
)
Removed temporary PoseMark directory: /tmp/posemark-76fx4td3
Total running time of the script: (0 minutes 4.365 seconds)