A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

Lightweight OpenPose — LiteRT Pose Estimation

Locate human body joints in an image for pose visualization. Returns 18 anonymous keypoints; it does not identify people or track them across video.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Locate human body joints in an image for pose visualization. Returns 18 anonymous keypoints; it does not identify people or track them across video.

Created by
Daniil Osokin
License
Apache-2.0
Model released
Unknown
Source repository created
Mar 2, 2019
Model weights · FP16
8.3 MB

Best for

  • Single-person pose visualization.
  • Gesture and exercise prototypes that do not require identity recognition.

Keep in mind

  • The simple argmax decoder returns one location per keypoint channel and is intended for one prominent person, not full multi-person association.
  • Occlusion, cropped bodies, unusual poses, blur, and small people reduce accuracy.
  • Coordinates are estimates and must not be used for medical, biometric, safety, or surveillance decisions.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
Daniil-Osokin/lightweight-human-pose-estimation.pytorch
Release verification
Unknown
Release evidence
Exact original checkpoint public release date has not been verified. Repository creation, training, paper, artifact conversion and update dates are not release evidence.
Source updated
Apr 30, 2024
Browser model files
litert-community/lightweight-openpose
Files updated
Sep 1, 2026
Artifact revision
7a1d619e67f829fc84c4b7a7e271384b5508762e

Input

IMAGE

Sample result

Lightweight OpenPose — LiteRT Pose Estimation

Settings

Adds a self-contained SVG presentation image alongside the unchanged native model fields. When enabled, draws model predictions or presents generated media and text. Adds CPU postprocessing time and image bytes. It does not change model predictions.
Advanced settings 1
Marks decoded keypoints below this heatmap confidence as not visible. Lower values retain more uncertain keypoints. Higher values hide weaker keypoints and may omit real joints; this filters visibility rather than improving model accuracy.

Output

Your result will appear here

Add input or try the sample, then run the model.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

Lightweight OpenPose — LiteRT Pose Estimation

4 configurations tested
Passed4
Failed0
TestedSep 8, 2026
RuntimeLiteRT.js 2.5.3

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResult
FP16GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedModerate
FP32CPU · WASMPassedFast
FP16CPU · WASMPassedModerate
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-09-08T17:48:00.000Z
Environment
Headless Chrome 151 on macOS; LiteRT.js 2.5.3
Runtime
LiteRT.js 2.5.3
Artifact revision
7a1d619e67f829fc84c4b7a7e271384b5508762e
Fixture
{
  "file": "modnet-portrait-pexels-5965592.jpg",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/modnet-portrait-pexels-5965592.jpg",
  "size": 78124,
  "sha1": "38212739a4a4f0be25d84863cf991d7fb6c6d818",
  "width": 1024,
  "height": 683,
  "license": "Pexels License"
}
Rerun when
model artifact revision change; LiteRT.js version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "grade": "reference",
      "visibleKeypoints": 4
    },
    "runtime": {
      "loadMs": 7723.7,
      "inferenceMs": 166.2,
      "totalMs": 7889.9,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "grade": "excellent",
      "matchingGridLocations": 18,
      "maximumConfidenceDifference": 0.000008
    },
    "runtime": {
      "loadMs": 9065.5,
      "inferenceMs": 536,
      "totalMs": 9601.5,
      "grade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "grade": "acceptable",
      "matchingGridLocations": 16,
      "maximumGridDistance": 0.0625
    },
    "runtime": {
      "loadMs": 6411.7,
      "inferenceMs": 714.8,
      "totalMs": 7126.5,
      "grade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "grade": "acceptable",
      "matchingGridLocations": 16,
      "maximumGridDistance": 0.0625
    },
    "runtime": {
      "loadMs": 6405.9,
      "inferenceMs": 50.9,
      "totalMs": 6456.8,
      "grade": "good"
    }
  }
]

Something not quite right?

Tell us what happened with Lightweight OpenPose. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.