Hugging Face's logo

qualcomm
/
MediaPipe-Face-Detection

Object Detection
PyTorch
TF Lite
ONNX
real_time
android
Model card Files Community
1
qaihm-bot commited on
Commit
a1f51b7
·
verified ·
1 Parent(s): 4587586

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +70 -28
README.md CHANGED
@@ -15,7 +15,7 @@ tags:
15
 
16
  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.
17
 
18
- This model is an implementation of MediaPipe-Face-Detection found [here](https://github.com/zmurez/MediaPipePyTorch/).
19
  This repository provides scripts to run MediaPipe-Face-Detection on Qualcomm® devices.
20
  More details on model performance across various devices, can be found
21
  [here](https://aihub.qualcomm.com/models/mediapipe_face).
@@ -32,17 +32,53 @@ More details on model performance across various devices, can be found
32
  - Model size (MediaPipeFaceLandmarkDetector): 2.34 MB
33
  - Number of output classes: 6
34
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
 
36
 
37
 
38
- | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
39
- | ---|---|---|---|---|---|---|---|
40
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | TFLite | 0.552 ms | 0 - 1 MB | FP16 | NPU | [MediaPipeFaceDetector.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite)
41
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | TFLite | 0.193 ms | 0 - 4 MB | FP16 | NPU | [MediaPipeFaceLandmarkDetector.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite)
42
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 0.622 ms | 0 - 5 MB | FP16 | NPU | [MediaPipeFaceDetector.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.so)
43
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 0.277 ms | 0 - 8 MB | FP16 | NPU | [MediaPipeFaceLandmarkDetector.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.so)
44
-
45
-
46
 
47
  ## Installation
48
 
@@ -97,23 +133,25 @@ device. This script does the following:
97
  ```bash
98
  python -m qai_hub_models.models.mediapipe_face.export
99
  ```
100
-
101
  ```
102
- Profile Job summary of MediaPipeFaceDetector
103
- --------------------------------------------------
104
- Device: Snapdragon X Elite CRD (11)
105
- Estimated Inference Time: 0.77 ms
106
- Estimated Peak Memory Range: 0.75-0.75 MB
107
- Compute Units: NPU (146) | Total (146)
108
-
109
- Profile Job summary of MediaPipeFaceLandmarkDetector
110
- --------------------------------------------------
111
- Device: Snapdragon X Elite CRD (11)
112
- Estimated Inference Time: 0.38 ms
113
- Estimated Peak Memory Range: 0.42-0.42 MB
114
- Compute Units: NPU (105) | Total (105)
115
-
116
-
 
 
 
117
  ```
118
 
119
 
@@ -241,15 +279,19 @@ provides instructions on how to use the `.so` shared library in an Android appl
241
  Get more details on MediaPipe-Face-Detection's performance across various devices [here](https://aihub.qualcomm.com/models/mediapipe_face).
242
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
243
 
 
244
  ## License
245
- - The license for the original implementation of MediaPipe-Face-Detection can be found
246
- [here](https://github.com/zmurez/MediaPipePyTorch/blob/master/LICENSE).
247
- - 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)
 
248
 
249
  ## References
250
  * [BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs](https://arxiv.org/abs/1907.05047)
251
  * [Source Model Implementation](https://github.com/zmurez/MediaPipePyTorch/)
252
 
 
 
253
  ## Community
254
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
255
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
 
15
 
16
  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.
17
 
18
+ This model is an implementation of MediaPipe-Face-Detection found [here]({source_repo}).
19
  This repository provides scripts to run MediaPipe-Face-Detection on Qualcomm® devices.
20
  More details on model performance across various devices, can be found
21
  [here](https://aihub.qualcomm.com/models/mediapipe_face).
 
32
  - Model size (MediaPipeFaceLandmarkDetector): 2.34 MB
33
  - Number of output classes: 6
34
 
35
+ | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
36
+ |---|---|---|---|---|---|---|---|---|
37
+ | MediaPipeFaceDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.549 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
38
+ | MediaPipeFaceDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.626 ms | 1 - 6 MB | FP16 | NPU | [MediaPipe-Face-Detection.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.so) |
39
+ | MediaPipeFaceDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.003 ms | 0 - 74 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.onnx) |
40
+ | MediaPipeFaceDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.45 ms | 0 - 34 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
41
+ | MediaPipeFaceDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.502 ms | 1 - 15 MB | FP16 | NPU | [MediaPipe-Face-Detection.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.so) |
42
+ | MediaPipeFaceDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 0.81 ms | 0 - 39 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.onnx) |
43
+ | MediaPipeFaceDetector | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.546 ms | 0 - 72 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
44
+ | MediaPipeFaceDetector | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.599 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
45
+ | MediaPipeFaceDetector | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.549 ms | 0 - 72 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
46
+ | MediaPipeFaceDetector | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.602 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
47
+ | MediaPipeFaceDetector | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 0.549 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
48
+ | MediaPipeFaceDetector | SA8775 (Proxy) | SA8775P Proxy | QNN | 0.612 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
49
+ | MediaPipeFaceDetector | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.547 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
50
+ | MediaPipeFaceDetector | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.613 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
51
+ | MediaPipeFaceDetector | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.763 ms | 0 - 31 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
52
+ | MediaPipeFaceDetector | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.83 ms | 1 - 17 MB | FP16 | NPU | Use Export Script |
53
+ | MediaPipeFaceDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.41 ms | 0 - 24 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.tflite) |
54
+ | MediaPipeFaceDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.38 ms | 1 - 13 MB | FP16 | NPU | Use Export Script |
55
+ | MediaPipeFaceDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 0.752 ms | 0 - 28 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.onnx) |
56
+ | MediaPipeFaceDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.804 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
57
+ | MediaPipeFaceDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.04 ms | 2 - 2 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceDetector.onnx) |
58
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.19 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
59
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.279 ms | 0 - 7 MB | FP16 | NPU | [MediaPipe-Face-Detection.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.so) |
60
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 0.506 ms | 0 - 2 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.onnx) |
61
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.144 ms | 0 - 29 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
62
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.213 ms | 0 - 12 MB | FP16 | NPU | [MediaPipe-Face-Detection.so](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.so) |
63
+ | MediaPipeFaceLandmarkDetector | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 0.402 ms | 0 - 32 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.onnx) |
64
+ | MediaPipeFaceLandmarkDetector | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.19 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
65
+ | MediaPipeFaceLandmarkDetector | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.274 ms | 0 - 2 MB | FP16 | NPU | Use Export Script |
66
+ | MediaPipeFaceLandmarkDetector | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.193 ms | 0 - 17 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
67
+ | MediaPipeFaceLandmarkDetector | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.275 ms | 0 - 2 MB | FP16 | NPU | Use Export Script |
68
+ | MediaPipeFaceLandmarkDetector | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 0.194 ms | 0 - 1 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
69
+ | MediaPipeFaceLandmarkDetector | SA8775 (Proxy) | SA8775P Proxy | QNN | 0.277 ms | 0 - 2 MB | FP16 | NPU | Use Export Script |
70
+ | MediaPipeFaceLandmarkDetector | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.194 ms | 0 - 74 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
71
+ | MediaPipeFaceLandmarkDetector | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.273 ms | 0 - 2 MB | FP16 | NPU | Use Export Script |
72
+ | MediaPipeFaceLandmarkDetector | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.283 ms | 0 - 29 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
73
+ | MediaPipeFaceLandmarkDetector | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.382 ms | 0 - 14 MB | FP16 | NPU | Use Export Script |
74
+ | MediaPipeFaceLandmarkDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.122 ms | 0 - 18 MB | FP16 | NPU | [MediaPipe-Face-Detection.tflite](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.tflite) |
75
+ | MediaPipeFaceLandmarkDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.207 ms | 0 - 10 MB | FP16 | NPU | Use Export Script |
76
+ | MediaPipeFaceLandmarkDetector | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 0.409 ms | 0 - 18 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.onnx) |
77
+ | MediaPipeFaceLandmarkDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.383 ms | 0 - 0 MB | FP16 | NPU | Use Export Script |
78
+ | MediaPipeFaceLandmarkDetector | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.512 ms | 2 - 2 MB | FP16 | NPU | [MediaPipe-Face-Detection.onnx](https://huggingface.co/qualcomm/MediaPipe-Face-Detection/blob/main/MediaPipeFaceLandmarkDetector.onnx) |
79
 
80
 
81
 
 
 
 
 
 
 
 
 
82
 
83
  ## Installation
84
 
 
133
  ```bash
134
  python -m qai_hub_models.models.mediapipe_face.export
135
  ```
 
136
  ```
137
+ Profiling Results
138
+ ------------------------------------------------------------
139
+ MediaPipeFaceDetector
140
+ Device : Samsung Galaxy S23 (13)
141
+ Runtime : TFLITE
142
+ Estimated inference time (ms) : 0.5
143
+ Estimated peak memory usage (MB): [0, 1]
144
+ Total # Ops : 111
145
+ Compute Unit(s) : NPU (111 ops)
146
+
147
+ ------------------------------------------------------------
148
+ MediaPipeFaceLandmarkDetector
149
+ Device : Samsung Galaxy S23 (13)
150
+ Runtime : TFLITE
151
+ Estimated inference time (ms) : 0.2
152
+ Estimated peak memory usage (MB): [0, 1]
153
+ Total # Ops : 100
154
+ Compute Unit(s) : NPU (100 ops)
155
  ```
156
 
157
 
 
279
  Get more details on MediaPipe-Face-Detection's performance across various devices [here](https://aihub.qualcomm.com/models/mediapipe_face).
280
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
281
 
282
+
283
  ## License
284
+ * The license for the original implementation of MediaPipe-Face-Detection can be found [here](https://github.com/zmurez/MediaPipePyTorch/blob/master/LICENSE).
285
+ * 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)
286
+
287
+
288
 
289
  ## References
290
  * [BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs](https://arxiv.org/abs/1907.05047)
291
  * [Source Model Implementation](https://github.com/zmurez/MediaPipePyTorch/)
292
 
293
+
294
+
295
  ## Community
296
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
297
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).