You can deploy this model and start earning money today!

tomas-gajarsky/
facetorch-align-synergynet

Name: facetorch-align-synergynet
Availability: InStock

face-alignment

facial-landmark-detection

2110.09772

mit

Views

No views yet

SynergyNet 3D face alignment

SynergyNet MobileNetV2 checkpoint for approximate 3D facial alignment.

This repository contains immutable model artifacts used by Facetorch. Use the packaged Facetorch manifest to select a revision and artifact; do not treat mutable main or older unlisted files as a release contract.

Contract

Field	Value
Model ID	`align-synergynet`
Architecture	SynergyNet with MobileNetV2 backbone
Input	120 by 120 RGB face crop
Output	62 values per face: 12 pose, 40 shape, and 10 expression coefficients.
Dynamic shapes	Batch dimension 1 through 64.
Weights license	MIT

Preprocessing: Resize the Facetorch face crop to 120 by 120; no additional normalization is configured.

Release artifacts

File	Format	Runtime	Devices	SHA-256
`model-torch2.6.pt2`	pt2	>=2.6, <2.7	cpu, cuda	`64cca00d6baa6d6ac37acb060a8e03bd5c5cab6467b9f1819e02a7ccb4a6b4b5`
`model-torch2.11.pt2`	pt2	>=2.11, <2.12	cpu, cuda	`ebaec95ba16cfee44277b32c34017ad751387fc115bf5da0ac458fa59027c21c`
`model.pt`	torchscript	>=2.6, <2.12	cpu	`09220491e3a0c11eb1c076b9c6551e214a078b54b86acb6f0e94165e00785df4`

Facetorch v1 supports the Torch 2.6 and 2.11 cohort files listed in its manifest. The legacy TorchScript object is CPU-only and requires the explicit legacy opt-in. Files from unsupported cohorts are not part of the v1 release contract.

Loading the manifest-selected artifact

python

import torch
from huggingface_hub import hf_hub_download
from facetorch.artifacts import get_model_manifest

MODEL_ID = "align-synergynet"
device = "cuda" if torch.cuda.is_available() else "cpu"
artifact = get_model_manifest().candidates(
    MODEL_ID,
    torch_version=torch.__version__,
    device=device,
    allow_legacy_models=False,
)[0]
path = hf_hub_download(
    repo_id=artifact.repo_id,
    revision=artifact.revision,
    filename=artifact.filename,
)
model = torch.export.load(path).module().to(device).eval()
example = torch.randn(1, 3, 120, 120, device=device)
with torch.inference_mode():
    output = model(example)

The random tensor above is only a loading smoke test. Use Facetorch's documented preprocessing for meaningful inference.

Provenance

Upstream	Immutable revision	Role	License
https://github.com/choyingw/SynergyNet	`9de11e2831b85254d776ca748fa2ffd68aedf4ba`	checkpoint publisher and architecture source	MIT

Upstream checkpoint	SHA-256	Source
`pretrained/best.pth.tar`	`128471cf9920833c2ef2e2868a6f6ae2f680acb1ac77274bffcf2c638053f42b`	publisher location

Mapping method: exact_tensor_equality.

Result: 445 of 445 model tensors matched exactly; upstream-only 3DMM texture tensors u_tex and w_tex are not model parameters.

The repository owner approved the mapping and redistribution record on 2026-08-23. Under the recorded policy, an author-published checkpoint in a permissively licensed repository with no separate checkpoint terms uses that repository license. MIT and Apache-2.0 have not been converted or treated as interchangeable. See LICENSE, THIRD_PARTY_NOTICES.md, and Facetorch's facetorch/models/governance.json.

Papers

Synergy between 3DMM and 3D Landmarks for Accurate 3D Facial Geometry

Intended use

Research and visualization of approximate 3D facial alignment.

Limitations and responsible use

Landmarks and pose are estimates and degrade with occlusion, extreme pose, and out-of-domain imagery.
Do not interpret geometry estimates as medical or identity evidence.
The checkpoint license does not grant rights to upstream training datasets.
The artifact license does not itself license training datasets, input data, or a deployment's processing of personal data.
Do not use model output as the sole basis for consequential decisions about a person.

tomas-gajarsky/facetorch-align-synergynet