Hugging Face's logo

qualcomm
/
Lightweight-Face-Detection

Object Detection
PyTorch
LiteRT
real_time
bu_iot
android
Model card Files
xet
Community

Lightweight-Face-Detection: Optimized for Mobile Deployment

Lightweight and efficient face detector

A small and accurate model for detecting bounding boxes for faces in images. This model's architecture was developed by Qualcomm. The model was trained by Qualcomm on a proprietary dataset of faces, but can be used on any image.

This repository provides scripts to run Lightweight-Face-Detection on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.object_detection
  • Model Stats:
    • Model checkpoint: qfd360_sl_model.pt
    • Inference latency: RealTime
    • Input resolution: 480x640
    • Number of parameters: 878K
    • Model size (float): 3.37 MB
    • Model size (w8a8): 965 KB
    • Model size (w8a16): 1.09 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
Lightweight-Face-Detection float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 7.589 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 6.639 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 3.506 ms 0 - 32 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 4.093 ms 1 - 32 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 3.062 ms 0 - 9 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 2.522 ms 1 - 8 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 2.694 ms 0 - 10 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 3.846 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 3.185 ms 1 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float SA7255P ADP Qualcomm® SA7255P TFLITE 7.589 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float SA7255P ADP Qualcomm® SA7255P QNN_DLC 6.639 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 3.065 ms 0 - 10 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 2.526 ms 1 - 8 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float SA8295P ADP Qualcomm® SA8295P TFLITE 4.293 ms 0 - 27 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float SA8295P ADP Qualcomm® SA8295P QNN_DLC 3.595 ms 1 - 28 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 3.067 ms 0 - 5 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 2.523 ms 1 - 8 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float SA8775P ADP Qualcomm® SA8775P TFLITE 3.846 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float SA8775P ADP Qualcomm® SA8775P QNN_DLC 3.185 ms 1 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 1.851 ms 0 - 31 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 1.569 ms 1 - 33 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 1.342 ms 0 - 34 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 1.548 ms 0 - 29 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 1.195 ms 1 - 27 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 1.208 ms 0 - 28 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 1.487 ms 0 - 27 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 1.106 ms 1 - 26 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 1.237 ms 1 - 31 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 2.783 ms 3 - 3 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 2.464 ms 1 - 1 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 5.759 ms 2 - 101 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 87.956 ms 28 - 32 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 3.048 ms 1 - 22 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 2.4 ms 1 - 31 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 1.694 ms 0 - 6 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 1.944 ms 0 - 10 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 2.026 ms 1 - 22 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 36.192 ms 26 - 28 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 SA7255P ADP Qualcomm® SA7255P QNN_DLC 3.048 ms 1 - 22 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 1.702 ms 1 - 6 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 SA8295P ADP Qualcomm® SA8295P QNN_DLC 2.265 ms 0 - 27 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 1.7 ms 0 - 7 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 SA8775P ADP Qualcomm® SA8775P QNN_DLC 2.026 ms 1 - 22 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.968 ms 1 - 33 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 1.13 ms 0 - 28 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.817 ms 1 - 29 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.951 ms 0 - 23 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 2.896 ms 1 - 28 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 41.579 ms 29 - 44 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.718 ms 1 - 29 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.933 ms 1 - 28 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a16 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1.914 ms 0 - 0 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.972 ms 1 - 1 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 TFLITE 1.335 ms 0 - 3 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 1.098 ms 2 - 102 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 17.021 ms 13 - 17 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 1.171 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 1.086 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 0.52 ms 0 - 31 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 0.501 ms 0 - 29 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.44 ms 0 - 8 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.417 ms 0 - 8 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 0.667 ms 0 - 9 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 0.674 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 0.624 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 9.364 ms 0 - 3 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 10.144 ms 11 - 15 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 1.171 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 1.086 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.444 ms 0 - 9 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.416 ms 0 - 4 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 0.868 ms 0 - 25 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 0.838 ms 0 - 26 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.443 ms 0 - 8 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.416 ms 0 - 8 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 0.674 ms 0 - 20 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 0.624 ms 0 - 20 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.304 ms 0 - 33 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.282 ms 0 - 35 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.434 ms 0 - 29 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.248 ms 0 - 21 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.235 ms 0 - 27 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.364 ms 0 - 23 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile TFLITE 0.54 ms 0 - 28 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 0.479 ms 0 - 22 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 11.639 ms 13 - 29 MB CPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.229 ms 0 - 23 MB NPU Lightweight-Face-Detection.tflite
Lightweight-Face-Detection w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.195 ms 0 - 27 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.331 ms 0 - 23 MB NPU Lightweight-Face-Detection.onnx.zip
Lightweight-Face-Detection w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.526 ms 0 - 0 MB NPU Lightweight-Face-Detection.dlc
Lightweight-Face-Detection w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.599 ms 0 - 0 MB NPU Lightweight-Face-Detection.onnx.zip

Installation

Install the package via pip:

# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install "qai-hub-models[face-det-lite]"

Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub Workbench with your Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.

With this API token, you can configure your client to run models on the cloud hosted devices.

qai-hub configure --api_token API_TOKEN

Navigate to docs for more information.

Demo off target

The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.

python -m qai_hub_models.models.face_det_lite.demo

The above demo runs a reference implementation of pre-processing, model inference, and post processing.

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.face_det_lite.demo

Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:

  • Performance check on-device on a cloud-hosted device
  • Downloads compiled assets that can be deployed on-device for Android.
  • Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.face_det_lite.export

How does this work?

This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:

Step 1: Compile model for on-device deployment

To compile a PyTorch model for on-device deployment, we first trace the model in memory using the jit.trace and then call the submit_compile_job API.

import torch

import qai_hub as hub
from qai_hub_models.models.face_det_lite import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S25")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

Step 2: Performance profiling on cloud-hosted device

After compiling models from step 1. Models can be profiled model on-device using the target_model. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics.

profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)

Step 3: Verify on-device accuracy

To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.

input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.

Note: This on-device profiling and inference requires access to Qualcomm® AI Hub Workbench. Sign up for access.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.face_det_lite.demo --eval-mode on-device

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.face_det_lite.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Lightweight-Face-Detection's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Lightweight-Face-Detection can be found here.
  • The license for the compiled assets for on-device deployment can be found here

Community

Downloads last month
933
Inference Providers NEW
Object Detection
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support