You can deploy this model and start earning money today!

takoyakisoft/
vtuberkit-vision-landmarkers-onnx

Name: vtuberkit-vision-landmarkers-onnx
Availability: InStock

onnx

apache-2.0

Views

No views yet

VTuberKit Vision Landmarkers ONNX

This public repository packages Face / Body / Hand vision landmark ONNX models for VTuberKitForYMM4.

Normal VTuberKitForYMM4 users do not need a Hugging Face token. The repository is public, publicRepo is true, and requiresAuth is false in the manifest.

Layout

onnx/face: face detector, face landmark, face blendshape
onnx/body: pose detector and pose landmark, stored as Body for VTuberKit local paths
onnx/hand: palm detector, hand recrop, and hand landmark
manifest/vtuberkit_vision_manifest.json: machine-readable VTuberKit manifest
metadata: per-model source, SHA256, size, license, and verification metadata
licenses: upstream licenses and notices
sources: source records for the original model bundles

Sources And Conversion

Face, Body pose landmark, and Hand models are converted from official Google MediaPipe .task bundles with the Tools/VisionOnnxExporter uv tool. The source license is Apache-2.0, allowing commercial use, modification, and redistribution.

The Body pose detector could not be converted from the official MediaPipe TFLite with the current converter stack because the model contains sparse/DENSIFY constants. For that model only, this package uses the OpenCV Zoo MediaPipe person detector ONNX fallback. Its source and Apache-2.0 license are recorded in metadata/body/pose_detector_conversion.json and metadata/body/pose_detector.json.

VTuberKitForYMM4 Local Paths

Face: MediaPipeToolForYMM4/Assets/Models/face
Body: MediaPipeToolForYMM4/Assets/Models/body
Hand: MediaPipeToolForYMM4/Assets/Models/hand

Manifest

Read manifest/vtuberkit_vision_manifest.json. Each model entry includes path, localPath, sha256, sizeBytes, source, sourceLicense, and commercialUseAllowed. The body modality has alias pose because the underlying model family is MediaPipe Pose.

ONNX SHA256

onnx/face/face_detector.onnx 0acb84ef882f18005675b14b3d8021038fab9fc2d2e6365fc73787022d7dda48
onnx/face/face_landmark.onnx 38ea4d75bffea5a72cb93ec027ddeaf31763049ffa57ec944fa3840119083bec
onnx/face/face_blendshape.onnx 4d97c181a04c1ac1778daa7ffbd9ae1e80b65300750078a7edbcf456d29b53c3
onnx/body/pose_detector.onnx 47fd5599d6fa17608f03e0eb0ae230baa6e597d7e8a2c8199fe00abea55a701f
onnx/body/pose_landmark.onnx eef49ab70a13e6a721ca936f40bd3c9c008058723a99012c769d98ea12416ab2
onnx/hand/palm_detector.onnx 939e53b8c438bdbdcaaf2e5589c8204cb6c92d383a6fe87c939ca3bb52f605c7
onnx/hand/hand_recrop.onnx d71150bd45f3a0060805bda7147302d3d9215a26e7bc270ef840f9718f55d797
onnx/hand/hand_landmark.onnx 716f2d75eb41cbb8f222cda9099eb617ec0a3248ecb96e77c0dc916759b5bb1c

takoyakisoft/vtuberkit-vision-landmarkers-onnx