Skip to content

Seeeeeyo/opencap-visualizer-pip

BranchesTags

Repository files navigation

OpenCap Visualizer

Generate videos from OpenCap biomechanics data files with both command-line interface and Python API.

This tool uses the deployed OpenCap Visualizer to create videos from biomechanics data files (.json, .osim/.mot pairs) using headless browser automation.

Features

  • Dual Interface: Both command-line tool and Python API
  • No Local Setup Required: Uses deployed web application by default
  • Multiple Data Formats: Supports JSON files and OpenSim .osim/.mot pairs
  • Subject Comparison: Generate videos with multiple subjects
  • Anatomical Camera Views: Use biomechanics-friendly camera angles
  • Customizable: Colors, zoom, centering, loops, and dimensions
  • Geometry model selection: Choose the S3 geometry set used for visualization (LaiArnold or Hu_ISB_shoulder), matching the web app’s model picker
  • Automatic 3D Geometry: Loads realistic human models from cloud storage
  • Realtime live IK WebSocket API: Build and send init / frame messages from Python (build_live_init_dict, send_live_frame, etc.) without a precomputed full-motion JSON

Installation

pip install opencap-visualizer

Note: After installation, you'll need to install browser dependencies:

playwright install chromium

Live IK streaming (WebSocket server)

Install the optional dependency:

pip install opencap-visualizer[live]

Realtime streaming (Python API)

For incremental data (IK or any pipeline that produces one pose per timestep), run your own WebSocket server and use the same protocol as the visualizer’s Live IK Stream panel:

  1. build_live_init_dict — skeleton metadata (subjects with per-body attachedGeometries / scaleFactors), frameRate, optional camera / model / styles (same options as file replay).
  2. send_live_init(websocket, init_dict, mesh_load_delay=...) — send init; use mesh_load_delay=0 for minimal latency after meshes load.
  3. send_live_frame(websocket, streams) — send one timestep; streams is { subject_id: { "time": float, "bodies": { bone: { "rotation": [...], "translation": [...] } } } }.
  4. broadcast_live_frame(streams) — same as send_live_frame but to every connected client (like send_notification).
  5. skeleton_bodies_from_visualizer_json(path) — load bone metadata from any exported visualizer JSON to avoid hand-copying geometry lists.
import asyncio
import websockets
from opencap_visualizer import (
    build_live_init_dict,
    send_live_init,
    send_live_frame,
    skeleton_bodies_from_visualizer_json,
)

TEMPLATE_JSON = "/path/to/one_time_export.json"  # visualizer export with correct bodies/geometries

async def handler(websocket):
    bodies_meta = skeleton_bodies_from_visualizer_json(TEMPLATE_JSON)
    init_dict = build_live_init_dict(
        [{"id": "ik", "label": "Live", "bodies": bodies_meta, "model": "LaiArnold"}],
        frame_rate=60.0,
        camera="anterior",
    )
    await send_live_init(websocket, init_dict, mesh_load_delay=0.0)
    t = 0.0
    dt = 1.0 / 60.0
    while True:
        # Build bodies from your solver: each bone needs rotation [rx,ry,rz] and translation [x,y,z]
        bodies_pose = compute_pose_from_your_ik()  # dict[str, dict]
        await send_live_frame(
            websocket,
            {"ik": {"time": t, "bodies": bodies_pose}},
        )
        t += dt
        await asyncio.sleep(dt)

async def main():
    async with websockets.serve(handler, "0.0.0.0", 8765):
        await asyncio.Future()  # run forever

# asyncio.run(main())

In the browser, connect to ws://localhost:8765, then press Play on the timeline so new frames update the pose (the client may start paused after init).

See TESTING.md for smoke tests, examples/realtime_live_server.py, and a minimal WebSocket client check.

File replay (CLI / testing)

To replay a full visualizer JSON file (complete time series on disk):

opencap-visualizer-stream subject.json
# or: opencap-viz-stream subject1.json subject2.json 2.0

If port 8765 is already in use, use another port and the same port in the visualizer:

python -m opencap_visualizer.live_stream --port 8766 subject.json

Use a real path to your JSON (example asset: opencap-visualizer/public/samples/walk/sample_mono.json in the web app repo). python -m opencap_visualizer.live_stream matches the main visualizer repo’s live_stream_from_json.py.

Live stream CLI reference (file replay)

Positional Description
FILE.json One or two visualizer JSON files. Subject IDs in the protocol are usually the file stems (e.g. trial.json → trial).
speed Optional number after the JSON paths. Playback uses timestamps in the JSON; 1.0 ≈ real time, 2.0 ≈ 2× faster, 0.5 ≈ half speed. Default 1.0. Parsed as the last non-.json argument that remains after flags.
Option Description
--port N Listen port (default 8765). Use if you see “address already in use”; set the same port in the visualizer WebSocket URL.
--camera … Initial camera: anatomical preset (anterior, posterior, sagittal_right, sagittal_left, superior, inferior), axis / corner preset (front, isometric, frontTopRight, …), or inline JSON {"position":[x,y,z],"target":[x,y,z]}.
--model … Geometry folder name(s) matching the web app (e.g. LaiArnold). One value for all subjects, or comma-separated per subject.
--subject-colors Comma-separated hex colors, one per subject, e.g. "#d3d3d3,#4995e0". Uniform color per subject (overrides per-bone color from --body-style for those subjects).
--subject-opacity Comma-separated floats 0–1, one per subject (whole-model transparency).
--body-style Path to a JSON file, or inline JSON: a single object applies the same per-bone map to every subject; a JSON array gives one object per subject. Bones use { "color": "#rrggbb", "visible": true/false, … } as in the visualizer.

Flag order is flexible (--port 8766 --camera anterior subject.json 50 is fine). The server binds to 0.0.0.0 (all interfaces) so other machines on the LAN can use ws://<your-ip>:<port>.

Interactive stdin (terminal where the server runs; skipped if stdin is not available):

Command Effect
notify … / notify success … Banner on connected clients (info / success / warning / error).
camera … Update camera without reloading subjects (same forms as --camera).
scores n1 … n5 … Trial scores overlay; optional labels, colors: g o r …, title: ….
hidescores Hide scores overlay.
hide SUBJECT_ID / show SUBJECT_ID Subject visibility (IDs printed at connect, e.g. file stem).
help List commands.

Command Line Usage

Basic Examples

# Single subject
opencap-visualizer data.json -o output.mp4

# Multiple subjects comparison
opencap-visualizer subject1.json subject2.json -o comparison.mp4

# With custom settings
opencap-visualizer data.json --camera anterior --colors red --loops 2 -o front_view.mp4

# OpenSim files
opencap-visualizer model.osim motion.mot -o simulation.mp4

# Geometry model (must match the visualizer web app options)
opencap-visualizer data.json --model Hu_ISB_shoulder -o shoulder_model.mp4

Advanced Examples

# Multiple subjects with different colors
opencap-visualizer s1.json s2.json s3.json --colors red green blue --camera sagittal -o side_comparison.mp4

# High-resolution with custom zoom
opencap-visualizer data.json --width 3840 --height 2160 --zoom 0.8 --camera superior -o 4k_top_view.mp4

# Interactive mode for manual exploration
opencap-visualizer data.json --interactive --camera anterior

Camera Views

Anatomical Views (Recommended)

  • anterior / frontal / coronal - Front-facing view
  • posterior - Back view
  • sagittal / lateral - Side profile view
  • superior - Top-down view
  • inferior - Bottom-up view

Technical Views

  • top, bottom, front, back, left, right
  • isometric, default
  • Corner views: frontTopRight, backBottomLeft, etc.

Geometry model (--model)

The web visualizer resolves mesh geometry from named folders on cloud storage. In headless mode, pick the model before loading data (same choices as the app’s “Select Model” dialog):

Value Description
LaiArnold Default Lai Arnold full-body model
Hu_ISB_shoulder Hu ISB shoulder model 
opencap-visualizer trial.json --model LaiArnold -o out.mp4
opencap-visualizer model.osim motion.mot --model Hu_ISB_shoulder -o out.mp4

Command Line Options

positional arguments:
  FILE                  Data files (.json, or .osim/.mot pairs)

optional arguments:
  -o, --output PATH     Output video file (default: animation_video.mp4)
  --camera VIEW         Camera preset (anatomical or technical; optional)
  --model MODEL         Geometry folder: LaiArnold (default) or Hu_ISB_shoulder
  --colors COLOR...     Subject colors (hex or names: red, blue, #ff0000)
  --loops N             Animation loops to record (default: 1)
  --width PX            Video width (default: 1920)
  --height PX           Video height (default: 1080)
  --zoom FACTOR         Zoom factor (>1.0 = zoom out, default: 1.5)
  --no-center           Disable auto-centering on subjects
  --timeout SEC         Loading timeout in seconds (default: 120)
  --interactive         Open browser for manual exploration
  --vue-app-path PATH   Custom Vue app index.html path
  --dev-server-url URL  Custom Vue development server URL
  -v, --verbose         Enable verbose output

Python API Usage

Basic Examples

import opencap_visualizer as ocv

# Simple usage
success = ocv.create_video("data.json", "output.mp4")
if success:
    print("Video generated successfully!")

# Multiple subjects with settings
success = ocv.create_video(
    ["subject1.json", "subject2.json"],
    "comparison.mp4", 
    camera="anterior",
    model="LaiArnold",
    colors=["red", "blue"],
    loops=2,
    verbose=True
)

Class-based Usage

import opencap_visualizer as ocv

# Create visualizer instance
visualizer = ocv.OpenCapVisualizer(verbose=True)

# Generate video synchronously
success = visualizer.generate_video_sync(
    input_files=["subject1.json", "subject2.json"],
    output_path="comparison.mp4",
    camera="sagittal",
    model="Hu_ISB_shoulder",
    colors=["#ff0000", "#00ff00"],
    width=1920,
    height=1080,
    zoom=1.2
)

print(f"Success: {success}")

Async Usage

import asyncio
import opencap_visualizer as ocv

async def generate_videos():
    # Using convenience function
    success = await ocv.create_video_async(
        "data.json", 
        "output.mp4",
        camera="anterior",
        model="LaiArnold",
        colors=["blue"]
    )
    
    # Using class
    visualizer = ocv.OpenCapVisualizer(verbose=True)
    success = await visualizer.generate_video(
        ["s1.json", "s2.json", "s3.json"],
        "triple_comparison.mp4",
        camera="posterior",
        model="LaiArnold",
        colors=["red", "green", "blue"],
        center_subjects=True,
        zoom=1.5
    )
    
    return success

# Run async function
success = asyncio.run(generate_videos())

Live stream control helpers (Python)

With opencap-visualizer[live] and clients registered on your WebSocket server (as with the packaged CLI), you can broadcast UI commands to every connected browser:

import asyncio
from opencap_visualizer import send_notification, send_camera

async def ping_clients():
    await send_notification("Stream ready", level="success")
    await send_camera("anterior")

asyncio.run(ping_clients())

See opencap_visualizer.live_stream for stream_from_json (file replay), send_trial_scores, subject visibility, etc. Realtime init / frame helpers are documented under Realtime streaming above.

API Reference

OpenCapVisualizer Class

class OpenCapVisualizer:
    def __init__(self, verbose: bool = False)
    
    async def generate_video(
        self,
        input_files: Union[str, List[str]],
        output_path: str = "animation_video.mp4",
        *,
        vue_app_path: Optional[str] = None,
        dev_server_url: Optional[str] = None,
        width: int = 1920,
        height: int = 1080,
        timeout_seconds: int = 120,
        loops: int = 1,
        camera: Optional[str] = None,
        center_subjects: bool = True,
        zoom: float = 0.8,
        colors: Optional[List[str]] = None,
        model: str = "LaiArnold",
        interactive: bool = False
    ) -> bool
    
    def generate_video_sync(self, ...) -> bool  # Synchronous wrapper

Convenience Functions

async def create_video_async(input_files, output_path, **kwargs) -> bool
def create_video(input_files, output_path, **kwargs) -> bool

Data Formats

JSON Files

The tool accepts biomechanics JSON files with the following structure:

{
  "Data": {
    "ModelScalingVars": "path/to/model.osim",
    "Results": "path/to/motion.mot",
    "FrameTimesOG": [0.0, 0.033, 0.066, ...]
  }
}

OpenSim Files

Alternatively, provide .osim (model) and .mot (motion) file pairs:

opencap-visualizer model.osim motion.mot -o output.mp4

Dependencies

The tool automatically detects the best available option:

  1. Deployed Version (Recommended): https://opencap-visualizer.onrender.com/

    • No local setup required
    • Always up-to-date
    • Requires internet connection

  2. Local Development Server: http://localhost:3000

    • Start with npm run serve in the Vue.js project
    • Faster for development/testing

  3. Built Files: Local dist/index.html

    • Build with npm run build in the Vue.js project
    • Works offline

Configuration

Custom Servers

# Use local development server
opencap-visualizer data.json --dev-server-url http://localhost:3000

# Use custom built files
opencap-visualizer data.json --vue-app-path /path/to/dist/index.html

Environment Variables

# Set default development server
export OPENCAP_DEV_SERVER=http://localhost:3000

Troubleshooting

Browser Installation

If you get browser-related errors:

playwright install chromium

Connection Issues

  • Check internet connection for deployed version
  • For local development: npm run serve in Vue project
  • For built files: npm run build in Vue project

File Format Issues

  • Ensure JSON files contain valid biomechanics data structure
  • For OpenSim files, provide both .osim and .mot files
  • Check file paths are correct and accessible

Video Generation Issues

  • Increase timeout: --timeout 300
  • Enable verbose mode: --verbose or verbose=True
  • Try interactive mode: --interactive

License

MIT License - see LICENSE file for details.

Contributing

Contributions welcome! Please see the source repository for guidelines.

Support

For issues and questions:

About

No description, website, or topics provided.

Resources

Readme

License

MIT license
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

1 tags

Packages

 
 
 

Contributors

Languages