--- library_name: pytorch license: apache-2.0 pipeline_tag: object-detection tags: - real_time - quantized - android --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_face_quantized/web-assets/model_demo.png) # MediaPipe-Face-Detection-Quantized: Optimized for Mobile Deployment ## Detect faces and locate facial features in real-time video and image streams Designed for sub-millisecond processing, this model predicts bounding boxes and pose skeletons (left eye, right eye, nose tip, mouth, left eye tragion, and right eye tragion) of faces in an image. This model is an implementation of MediaPipe-Face-Detection-Quantized found [here](https://github.com/zmurez/MediaPipePyTorch/). This repository provides scripts to run MediaPipe-Face-Detection-Quantized on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/mediapipe_face_quantized). ### Model Details - **Model Type:** Object detection - **Model Stats:** - Input resolution: 256x256 - Number of output classes: 6 - Number of parameters (MediaPipeFaceDetector): 135K - Model size (MediaPipeFaceDetector): 255 KB - Number of parameters (MediaPipeFaceLandmarkDetector): 603K - Model size (MediaPipeFaceLandmarkDetector): 746 KB | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | MediaPipeFaceDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.279 ms | 0 - 25 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.303 ms | 0 - 79 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.so) | | MediaPipeFaceDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.185 ms | 0 - 16 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.21 ms | 0 - 15 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.so) | | MediaPipeFaceDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.201 ms | 0 - 19 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.176 ms | 0 - 15 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 0.685 ms | 0 - 22 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 0.74 ms | 0 - 7 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 4.998 ms | 0 - 4 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.278 ms | 0 - 10 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.305 ms | 0 - 1 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | SA7255P ADP | SA7255P | TFLITE | 2.155 ms | 0 - 19 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | SA7255P ADP | SA7255P | QNN | 2.313 ms | 0 - 10 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.272 ms | 0 - 10 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.305 ms | 0 - 1 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | SA8295P ADP | SA8295P | TFLITE | 0.664 ms | 0 - 14 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | SA8295P ADP | SA8295P | QNN | 0.95 ms | 0 - 5 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.276 ms | 0 - 11 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.305 ms | 0 - 2 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | SA8775P ADP | SA8775P | TFLITE | 0.606 ms | 0 - 19 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | SA8775P ADP | SA8775P | QNN | 0.807 ms | 0 - 6 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.333 ms | 0 - 19 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceDetector.tflite) | | MediaPipeFaceDetector | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.347 ms | 0 - 18 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.408 ms | 0 - 0 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.18 ms | 0 - 76 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.224 ms | 0 - 3 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.so) | | MediaPipeFaceLandmarkDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.142 ms | 0 - 14 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.161 ms | 0 - 14 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.so) | | MediaPipeFaceLandmarkDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.145 ms | 0 - 10 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.167 ms | 0 - 10 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 0.411 ms | 0 - 12 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 0.497 ms | 0 - 7 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 3.071 ms | 0 - 2 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.184 ms | 0 - 4 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.215 ms | 0 - 2 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | SA7255P ADP | SA7255P | TFLITE | 0.983 ms | 0 - 13 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | SA7255P ADP | SA7255P | QNN | 1.247 ms | 0 - 10 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.179 ms | 0 - 3 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.221 ms | 0 - 2 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | SA8295P ADP | SA8295P | TFLITE | 0.497 ms | 0 - 9 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | SA8295P ADP | SA8295P | QNN | 0.595 ms | 0 - 6 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.185 ms | 0 - 4 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.217 ms | 0 - 1 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | SA8775P ADP | SA8775P | TFLITE | 0.455 ms | 0 - 13 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | SA8775P ADP | SA8775P | QNN | 0.664 ms | 0 - 6 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.214 ms | 0 - 16 MB | FP16 | NPU | [MediaPipe-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection-Quantized/blob/main/MediaPipeFaceLandmarkDetector.tflite) | | MediaPipeFaceLandmarkDetector | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.267 ms | 0 - 13 MB | FP16 | NPU | Use Export Script | | MediaPipeFaceLandmarkDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.317 ms | 1 - 1 MB | FP16 | NPU | Use Export Script | ## Installation This model can be installed as a Python package via pip. ```bash pip install "qai-hub-models[mediapipe_face_quantized]" ``` ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) 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. ```bash qai-hub configure --api_token API_TOKEN ``` Navigate to [docs](https://app.aihub.qualcomm.com/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. ```bash python -m qai_hub_models.models.mediapipe_face_quantized.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.mediapipe_face_quantized.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. ```bash python -m qai_hub_models.models.mediapipe_face_quantized.export ``` ``` Profiling Results ------------------------------------------------------------ MediaPipeFaceDetector Device : Samsung Galaxy S23 (13) Runtime : TFLITE Estimated inference time (ms) : 0.3 Estimated peak memory usage (MB): [0, 25] Total # Ops : 121 Compute Unit(s) : NPU (121 ops) ------------------------------------------------------------ MediaPipeFaceLandmarkDetector Device : Samsung Galaxy S23 (13) Runtime : TFLITE Estimated inference time (ms) : 0.2 Estimated peak memory usage (MB): [0, 76] Total # Ops : 117 Compute Unit(s) : NPU (117 ops) ``` ## Deploying compiled model to Android The models can be deployed using multiple runtimes: - TensorFlow Lite (`.tflite` export): [This tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a guide to deploy the .tflite model in an Android application. - QNN (`.so` export ): This [sample app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html) provides instructions on how to use the `.so` shared library in an Android application. ## View on Qualcomm® AI Hub Get more details on MediaPipe-Face-Detection-Quantized's performance across various devices [here](https://aihub.qualcomm.com/models/mediapipe_face_quantized). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of MediaPipe-Face-Detection-Quantized can be found [here](https://github.com/zmurez/MediaPipePyTorch/blob/master/LICENSE). * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf) ## References * [BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs](https://arxiv.org/abs/1907.05047) * [Source Model Implementation](https://github.com/zmurez/MediaPipePyTorch/) ## Community * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI. * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).