Hugging Face's logo

qualcomm
/
Facial-Attribute-Detection

Object Detection
PyTorch
LiteRT
real_time
android
Model card Files Community
1
Facial-Attribute-Detection / README.md
qaihm-bot's picture
qaihm-bot
v0.36.0
24b7030 verified
preview code
|
raw
history blame contribute delete
23.1 kB
metadata
library_name: pytorch
license: other
tags:
  - real_time
  - android
pipeline_tag: object-detection

Facial-Attribute-Detection: Optimized for Mobile Deployment

Comprehensive facial analysis by extracting face features

Detects attributes (liveness, eye closeness, mask presence, glasses presence, sunglasses presence) that apply to a given face. 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 Facial-Attribute-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: multitask_FR_state_dict.pt
    • Input resolution: 128x128
    • Number of parameters: 12.1M
    • Model size (float): 46.3 MB
    • Model size (w8a8): 12.3 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
Facial-Attribute-Detection float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 4.383 ms 0 - 39 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 4.385 ms 0 - 25 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.195 ms 0 - 49 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 1.415 ms 0 - 34 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.873 ms 0 - 134 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.927 ms 0 - 10 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.394 ms 0 - 39 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1.479 ms 0 - 25 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float SA7255P ADP Qualcomm® SA7255P TFLITE 4.383 ms 0 - 39 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float SA7255P ADP Qualcomm® SA7255P QNN_DLC 4.385 ms 0 - 25 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.88 ms 0 - 134 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.919 ms 0 - 19 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float SA8295P ADP Qualcomm® SA8295P TFLITE 1.533 ms 0 - 44 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float SA8295P ADP Qualcomm® SA8295P QNN_DLC 1.611 ms 0 - 30 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.882 ms 0 - 148 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.918 ms 0 - 19 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float SA8775P ADP Qualcomm® SA8775P TFLITE 1.394 ms 0 - 39 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float SA8775P ADP Qualcomm® SA8775P QNN_DLC 1.479 ms 0 - 25 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 0.887 ms 0 - 138 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 0.92 ms 0 - 39 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 1.058 ms 0 - 74 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.677 ms 0 - 48 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.696 ms 0 - 32 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.832 ms 0 - 36 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 0.625 ms 0 - 44 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 0.63 ms 0 - 31 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 0.702 ms 0 - 26 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1.043 ms 82 - 82 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.069 ms 25 - 25 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 1.194 ms 0 - 35 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 1.129 ms 0 - 36 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 0.553 ms 0 - 54 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 0.657 ms 0 - 47 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.415 ms 0 - 64 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.407 ms 0 - 63 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 0.651 ms 0 - 35 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 0.628 ms 0 - 36 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) TFLITE 1.418 ms 0 - 46 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) QNN_DLC 1.596 ms 0 - 44 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 70.19 ms 2 - 5 MB CPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 1.194 ms 0 - 35 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 1.129 ms 0 - 36 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.407 ms 0 - 64 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.401 ms 0 - 61 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 0.885 ms 0 - 42 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 0.804 ms 0 - 42 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.42 ms 0 - 64 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.409 ms 0 - 51 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 0.651 ms 0 - 35 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 0.628 ms 0 - 36 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 0.42 ms 0 - 64 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 0.406 ms 0 - 11 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 0.588 ms 0 - 62 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.312 ms 0 - 56 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.301 ms 0 - 49 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.409 ms 0 - 57 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 0.307 ms 0 - 43 MB NPU Facial-Attribute-Detection.tflite
Facial-Attribute-Detection w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 0.247 ms 0 - 46 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 0.447 ms 0 - 39 MB NPU Facial-Attribute-Detection.onnx.zip
Facial-Attribute-Detection w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.53 ms 58 - 58 MB NPU Facial-Attribute-Detection.dlc
Facial-Attribute-Detection w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.615 ms 13 - 13 MB NPU Facial-Attribute-Detection.onnx.zip

Installation

Install the package via pip:

pip install qai-hub-models

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

Sign-in to Qualcomm® AI Hub 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_attrib_net.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_attrib_net.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_attrib_net.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_attrib_net import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S24")

# 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. 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_attrib_net.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_attrib_net.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 Facial-Attribute-Detection's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

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

Community