nwb-video-widgets 0.1.0

Interactive Jupyter widgets for NWB video and pose visualization

nwb-video-widgets

Interactive Jupyter widgets for NWB video and pose estimation visualization. Built with anywidget for compatibility across JupyterLab, Jupyter Notebook, VS Code, and Google Colab.

Table of Contents

• Installation
• Video Player Widgets
• Pose Estimation Widgets

Installation

For local only NWB file usage:

pip install nwb-video-widgets

For DANDI integration and streaming support:

pip install nwb-video-widgets[dandi]

Video Player Widgets

Multi-camera synchronized video player with configurable layout (Row, Column, or Grid).

Features:

• Interactive settings panel for video selection
• Multiple layout modes (Row, Column, Grid)
• Synchronized playback across all videos
• Session time display with NWB timestamps

DANDI Streaming

Use NWBDANDIVideoPlayer for videos hosted on DANDI:

from dandi.dandiapi import DandiAPIClient
from nwb_video_widgets import NWBDANDIVideoPlayer

client = DandiAPIClient()
dandiset = client.get_dandiset("000409", "draft")
asset = dandiset.get_asset_by_path("sub-NYU-39/sub-NYU-39_ses-..._behavior.nwb")

widget = NWBDANDIVideoPlayer(asset=asset)
widget

Local Files

Use NWBLocalVideoPlayer for local NWB files:

from pynwb import read_nwb
from nwb_video_widgets import NWBLocalVideoPlayer

nwbfile = read_nwb("experiment.nwb")
widget = NWBLocalVideoPlayer(nwbfile)
widget

---

Pose Estimation Widgets

Overlays DeepLabCut keypoints on streaming video with support for camera selection.

Features:

• Camera selection via settings panel
• Keypoint visibility toggles (All/None/individual)
• Label display toggle
• Session time display (NWB timestamps)
• Custom keypoint colors via colormap or explicit hex values
• Supports split files (videos in raw file, pose in processed file)

DANDI Streaming

Use NWBDANDIPoseEstimationWidget for DANDI-hosted files:

from dandi.dandiapi import DandiAPIClient
from nwb_video_widgets import NWBDANDIPoseEstimationWidget

client = DandiAPIClient()
dandiset = client.get_dandiset("000409", "draft")

# Single file (videos + pose in same file)
asset = dandiset.get_asset_by_path("sub-.../sub-..._combined.nwb")
widget = NWBDANDIPoseEstimationWidget(asset=asset)

# Or split files (videos in raw, pose in processed)
raw_asset = dandiset.get_asset_by_path("sub-.../sub-..._desc-raw.nwb")
processed_asset = dandiset.get_asset_by_path("sub-.../sub-..._desc-processed.nwb")
widget = NWBDANDIPoseEstimationWidget(
    asset=processed_asset,
    video_asset=raw_asset,
)
widget

Local Files

Use NWBLocalPoseEstimationWidget for local NWB files:

from pynwb import read_nwb
from nwb_video_widgets import NWBLocalPoseEstimationWidget

# Single file
nwbfile = read_nwb("experiment.nwb")
widget = NWBLocalPoseEstimationWidget(nwbfile)
widget

# Or split files
nwbfile_raw = read_nwb("raw.nwb")
nwbfile_processed = read_nwb("processed.nwb")
widget = NWBLocalPoseEstimationWidget(
    nwbfile=nwbfile_processed,
    video_nwbfile=nwbfile_raw,
)
widget

Parameters:

Parameter	Type	Description
keypoint_colors	str or dict	Matplotlib colormap name (e.g., 'tab10') or dict mapping keypoint names to hex colors
default_camera	str	Camera to display initially