metadata
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: object-detection
MediaPipe-Face-Detection: 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 found here.
This repository provides scripts to run MediaPipe-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:
- Input resolution: 256x256
- Number of output classes: 6
- Number of parameters (FaceDetector): 135K
- Model size (FaceDetector) (w8a8): 260 KB
- Number of parameters (FaceLandmarkDetector): 603K
- Model size (FaceLandmarkDetector) (w8a8): 750 KB
- Model size (FaceDetector) (float): 557 KB
- Model size (FaceLandmarkDetector) (float): 2.33 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| FaceDetector | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 2.937 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 2.853 ms | 1 - 20 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.759 ms | 0 - 32 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.745 ms | 1 - 34 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.543 ms | 0 - 6 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.535 ms | 1 - 7 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1.056 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 1.044 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 2.937 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 2.853 ms | 1 - 20 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.551 ms | 0 - 6 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.541 ms | 1 - 6 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 1.159 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 1.062 ms | 1 - 27 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.55 ms | 0 - 6 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.537 ms | 1 - 7 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1.056 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 1.044 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.551 ms | 0 - 7 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 0.544 ms | 1 - 7 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.837 ms | 0 - 7 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.395 ms | 0 - 28 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.38 ms | 1 - 28 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.583 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.408 ms | 0 - 25 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 0.388 ms | 0 - 22 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.589 ms | 0 - 20 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.693 ms | 5 - 5 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.886 ms | 2 - 2 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.834 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 0.852 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.27 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.331 ms | 0 - 27 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.185 ms | 0 - 10 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.214 ms | 0 - 10 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.436 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 0.437 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.834 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 0.852 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.197 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.213 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.557 ms | 0 - 22 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 0.564 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.195 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.215 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.436 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 0.437 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.197 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 0.21 ms | 0 - 9 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.374 ms | 0 - 8 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.147 ms | 0 - 27 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.155 ms | 0 - 28 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.262 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.185 ms | 0 - 24 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 0.189 ms | 0 - 22 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.305 ms | 0 - 19 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.325 ms | 8 - 8 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.402 ms | 2 - 2 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.663 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 0.649 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.267 ms | 0 - 35 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.281 ms | 0 - 34 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.234 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.234 ms | 0 - 12 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.441 ms | 0 - 20 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 0.442 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 0.641 ms | 0 - 22 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 0.636 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 5.038 ms | 0 - 3 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.663 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 0.649 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.236 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.234 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.565 ms | 0 - 29 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 0.557 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.233 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.234 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.441 ms | 0 - 20 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 0.442 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.238 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 0.239 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.808 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.16 ms | 0 - 35 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.148 ms | 0 - 33 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.562 ms | 0 - 38 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.159 ms | 0 - 27 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 0.154 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.494 ms | 0 - 30 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceDetector | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.351 ms | 0 - 0 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceDetector | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.967 ms | 0 - 0 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.483 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 0.463 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.195 ms | 0 - 30 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.194 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.154 ms | 0 - 12 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.15 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.359 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 0.323 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 0.386 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 0.381 ms | 0 - 21 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 2.923 ms | 0 - 3 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.483 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 0.463 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.159 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.153 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.451 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 0.418 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.159 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.153 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.359 ms | 0 - 16 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 0.323 ms | 0 - 17 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.152 ms | 0 - 11 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 0.154 ms | 0 - 12 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.312 ms | 0 - 12 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.118 ms | 0 - 26 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.104 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.21 ms | 0 - 27 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.139 ms | 0 - 24 MB | NPU | MediaPipe-Face-Detection.tflite |
| FaceLandmarkDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 0.121 ms | 0 - 23 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.255 ms | 0 - 22 MB | NPU | MediaPipe-Face-Detection.onnx.zip |
| FaceLandmarkDetector | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.239 ms | 3 - 3 MB | NPU | MediaPipe-Face-Detection.dlc |
| FaceLandmarkDetector | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.311 ms | 0 - 0 MB | NPU | MediaPipe-Face-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.mediapipe_face.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.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.mediapipe_face.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.mediapipe_face 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.
Deploying compiled model to Android
The models can be deployed using multiple runtimes:
TensorFlow Lite (
.tfliteexport): This tutorial provides a guide to deploy the .tflite model in an Android application.QNN (
.soexport ): This sample app provides instructions on how to use the.soshared library in an Android application.
View on Qualcomm® AI Hub
Get more details on MediaPipe-Face-Detection's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of MediaPipe-Face-Detection can be found here.
- The license for the compiled assets for on-device deployment can be found here
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.