---
family: yunet
license: bsd-3-clause
model_id: yunet-face
source_url: https://huggingface.co/litert-community/YuNet-Face-LiteRT
task: object-detection
---

# yunet-face

| | |
|---|---|
| **Task** | object-detection |
| **Family** | yunet |
| **Source** | https://huggingface.co/litert-community/YuNet-Face-LiteRT |
| **License** | bsd-3-clause |

## Conversion

| | |
|---|---|
| **Tool** | litert-torch 0.10.0 (editable dev checkout 115a136 + local patches) |
| **Command** | `python yunet/scripts/build_yunet.py` |
| **Quantization** | fp16 |

## Artifacts

| File | SHA-256 | Size (MB) |
|---|---|---|
| `5661d44b4bac479f17576dd5b2c14554e7924ab4d273ae14ddd6482b3fb81cab.tflite` | `ced5f52bef6e76ad4a66d1055b2b404336ceafbae8eeac8fed6aa9c7b2e4d776` | 0.244 |

## Performance

No benchmark data yet.

## Browser (LiteRT.js)

Model-level sweep results — measured behavior of this exact `.tflite` in the browser, only meaningful together with the environment that produced it. Full records (failure class, error evidence, sweep config): `data/sweep/2.5.3/2026-08-11/yunet-face.json`.

| Backend | Status | Full delegation | Output match | Max rel diff | Latency p50 (ms) | Environment | Date | Provenance |
|---|---|---|---|---|---|---|---|---|
| wasm_xnnpack | pass | - | - | - | 140.42 | mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 | 2026-08-11 | measured |
| webgpu_mldrift | pass | yes | pass | 0.013726315789473684 | 6.527 | mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 | 2026-08-11 | measured |

## Device runs (NPU / LiteRT-LM)

Model-level on-device results — measured behavior of this exact artifact on the recorded device, only meaningful together with that environment. Throughput is conditional on the measured prompt length (Prompt (tokens); '-' = the source stated none): rows differing only there are different measurements, not re-runs. Full records (failure classes, log evidence): `data/device_runs/2.2.0.dev20260804/2026-08-31/yunet-face__raspberry-pi-5.json`, `data/device_runs/2.2.0/2026-08-25/yunet-face__galaxy-s26.json`.

| Device | Accelerator | Status | Full delegation | Output match | Prompt (tokens) | Latency p50 (ms) | Prefill tok/s | Decode tok/s | TTFT (ms) | Peak mem (MB) | Environment | Date | Provenance |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| galaxy-s26 | gpu_mldrift | pass | - | - | - | 3.315 | - | - | - | - | Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 | 2026-08-25 | measured |
| galaxy-s26 | npu_qnn | pass | - | - | - | 1.85 | - | - | - | - | Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 | 2026-08-25 | measured |
| raspberry-pi-5 | cpu_xnnpack | pass | - | - | - | 33.696 | - | - | - | 137.5 | Raspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 | 2026-08-31 | measured |

## Pitfalls

- Input is BGR, 0-255, no normalization, letterboxed to 640x640; decode (priors + center/exp box + 5 landmarks + NMS over the 12 raw per-stride outputs) runs in the app (zoo README L1669-1675).

---

_Rendered by `edge-card` from `card.json` (the source of truth); edit the inputs, not this file._