Hugging Face's logo

qualcomm
/
MediaPipe-Pose-Estimation

Keypoint Detection
PyTorch
real_time
android
Model card Files
xet
Community
1
MediaPipe-Pose-Estimation
File size: 16,791 Bytes
da3e12a
 
0b50605
da3e12a
 
 
4a029e3
da3e12a
 
 
b42fd4e
da3e12a
2be3542
4aee478
130a74d
da3e12a
2be3542
 
 
 
 
 
 
 
 
 
 
 
 
 
2947060
 
 
 
 
 
2be3542
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2947060
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2be3542
2947060
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2be3542
2947060
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15dc2ba
da3e12a
2ce34f4
 
15dc2ba
da3e12a
 
 
 
 
9850f73
da3e12a
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
 
---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: keypoint-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/web-assets/model_demo.png)

# MediaPipe-Pose-Estimation: Optimized for Qualcomm Devices

The MediaPipe Pose Landmark Detector is a machine learning pipeline that predicts bounding boxes and pose skeletons of the face, hands, and torso in an image.

This is based on the implementation of MediaPipe-Pose-Estimation found [here](https://github.com/zmurez/MediaPipePyTorch/).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/mediapipe_pose) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mediapipe_pose/releases/v0.47.0/mediapipe_pose-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[MediaPipe-Pose-Estimation on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/mediapipe_pose)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/mediapipe_pose) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [MediaPipe-Pose-Estimation on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/mediapipe_pose) for usage instructions.

## Model Details

**Model Type:** Model_use_case.pose_estimation

**Model Stats:**
- Input resolution: 256x256
- Number of parameters (PoseDetector): 815K
- Model size (PoseDetector) (float): 3.14 MB
- Number of parameters (PoseLandmarkDetector): 3.36M
- Model size (PoseLandmarkDetector) (float): 12.9 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| PoseDetector | ONNX | float | Snapdragon® X Elite | 1.071 ms | 2 - 2 MB | NPU
| PoseDetector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.646 ms | 0 - 52 MB | NPU
| PoseDetector | ONNX | float | Qualcomm® QCS8550 (Proxy) | 0.896 ms | 0 - 5 MB | NPU
| PoseDetector | ONNX | float | Qualcomm® QCS9075 | 1.546 ms | 0 - 3 MB | NPU
| PoseDetector | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.537 ms | 0 - 36 MB | NPU
| PoseDetector | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.428 ms | 0 - 36 MB | NPU
| PoseDetector | ONNX | float | Snapdragon® X2 Elite | 0.519 ms | 0 - 0 MB | NPU
| PoseDetector | ONNX | w8a8 | Snapdragon® X Elite | 0.448 ms | 0 - 0 MB | NPU
| PoseDetector | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.24 ms | 0 - 45 MB | NPU
| PoseDetector | ONNX | w8a8 | Qualcomm® QCS6490 | 13.893 ms | 10 - 14 MB | CPU
| PoseDetector | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.337 ms | 0 - 22 MB | NPU
| PoseDetector | ONNX | w8a8 | Qualcomm® QCS9075 | 0.464 ms | 0 - 3 MB | NPU
| PoseDetector | ONNX | w8a8 | Qualcomm® QCM6690 | 10.463 ms | 8 - 17 MB | CPU
| PoseDetector | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.199 ms | 0 - 30 MB | NPU
| PoseDetector | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 7.954 ms | 10 - 19 MB | CPU
| PoseDetector | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.171 ms | 0 - 37 MB | NPU
| PoseDetector | ONNX | w8a8 | Snapdragon® X2 Elite | 0.19 ms | 0 - 0 MB | NPU
| PoseDetector | QNN_DLC | float | Snapdragon® X Elite | 1.031 ms | 0 - 0 MB | NPU
| PoseDetector | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.633 ms | 0 - 50 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 5.779 ms | 0 - 31 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 0.849 ms | 0 - 2 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® SA8775P | 1.596 ms | 0 - 35 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® QCS9075 | 1.485 ms | 0 - 2 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 2.155 ms | 0 - 43 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® SA7255P | 5.779 ms | 0 - 31 MB | NPU
| PoseDetector | QNN_DLC | float | Qualcomm® SA8295P | 2.541 ms | 0 - 23 MB | NPU
| PoseDetector | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.513 ms | 0 - 30 MB | NPU
| PoseDetector | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.395 ms | 0 - 34 MB | NPU
| PoseDetector | QNN_DLC | float | Snapdragon® X2 Elite | 0.638 ms | 0 - 0 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.43 ms | 0 - 0 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.23 ms | 0 - 41 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 1.548 ms | 0 - 2 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 0.99 ms | 0 - 28 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.31 ms | 0 - 2 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.507 ms | 0 - 30 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 0.402 ms | 0 - 2 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 3.227 ms | 0 - 142 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.567 ms | 0 - 42 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® SA7255P | 0.99 ms | 0 - 28 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.946 ms | 0 - 26 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.191 ms | 0 - 27 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.405 ms | 0 - 26 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.162 ms | 0 - 30 MB | NPU
| PoseDetector | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.287 ms | 0 - 0 MB | NPU
| PoseDetector | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.62 ms | 2 - 52 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 5.785 ms | 0 - 32 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 0.859 ms | 0 - 2 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® SA8775P | 6.994 ms | 0 - 32 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® QCS9075 | 1.489 ms | 0 - 5 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 2.165 ms | 0 - 44 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® SA7255P | 5.785 ms | 0 - 32 MB | NPU
| PoseDetector | TFLITE | float | Qualcomm® SA8295P | 2.554 ms | 0 - 25 MB | NPU
| PoseDetector | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.517 ms | 0 - 37 MB | NPU
| PoseDetector | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.395 ms | 0 - 36 MB | NPU
| PoseDetector | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.227 ms | 0 - 40 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCS6490 | 1.538 ms | 0 - 5 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 1.023 ms | 0 - 27 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.318 ms | 0 - 4 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® SA8775P | 1.906 ms | 0 - 28 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCS9075 | 0.409 ms | 0 - 3 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCM6690 | 3.189 ms | 0 - 141 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.573 ms | 0 - 42 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® SA7255P | 1.023 ms | 0 - 27 MB | NPU
| PoseDetector | TFLITE | w8a8 | Qualcomm® SA8295P | 0.928 ms | 0 - 25 MB | NPU
| PoseDetector | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.188 ms | 0 - 26 MB | NPU
| PoseDetector | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.406 ms | 0 - 25 MB | NPU
| PoseDetector | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.16 ms | 0 - 30 MB | NPU
| PoseLandmarkDetector | ONNX | float | Snapdragon® X Elite | 1.16 ms | 7 - 7 MB | NPU
| PoseLandmarkDetector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.663 ms | 0 - 88 MB | NPU
| PoseLandmarkDetector | ONNX | float | Qualcomm® QCS8550 (Proxy) | 0.903 ms | 0 - 10 MB | NPU
| PoseLandmarkDetector | ONNX | float | Qualcomm® QCS9075 | 1.367 ms | 1 - 3 MB | NPU
| PoseLandmarkDetector | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.552 ms | 0 - 52 MB | NPU
| PoseLandmarkDetector | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.459 ms | 0 - 51 MB | NPU
| PoseLandmarkDetector | ONNX | float | Snapdragon® X2 Elite | 0.582 ms | 7 - 7 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® X Elite | 0.498 ms | 4 - 4 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.236 ms | 0 - 66 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Qualcomm® QCS6490 | 18.746 ms | 8 - 15 MB | CPU
| PoseLandmarkDetector | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.347 ms | 0 - 9 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Qualcomm® QCS9075 | 0.523 ms | 0 - 3 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Qualcomm® QCM6690 | 9.906 ms | 0 - 12 MB | CPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.215 ms | 0 - 45 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 8.183 ms | 7 - 19 MB | CPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.179 ms | 0 - 54 MB | NPU
| PoseLandmarkDetector | ONNX | w8a8 | Snapdragon® X2 Elite | 0.19 ms | 0 - 0 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Snapdragon® X Elite | 1.065 ms | 1 - 1 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.61 ms | 0 - 77 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 3.162 ms | 1 - 41 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 0.817 ms | 1 - 2 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® SA8775P | 1.438 ms | 0 - 44 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® QCS9075 | 1.237 ms | 1 - 3 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1.189 ms | 0 - 69 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® SA7255P | 3.162 ms | 1 - 41 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Qualcomm® SA8295P | 1.411 ms | 0 - 32 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.5 ms | 1 - 45 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.432 ms | 1 - 45 MB | NPU
| PoseLandmarkDetector | QNN_DLC | float | Snapdragon® X2 Elite | 0.749 ms | 1 - 1 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.432 ms | 0 - 0 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.214 ms | 0 - 54 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 0.981 ms | 0 - 2 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 0.802 ms | 0 - 39 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.307 ms | 0 - 3 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.53 ms | 0 - 43 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 0.442 ms | 2 - 4 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 1.628 ms | 0 - 38 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.444 ms | 0 - 57 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® SA7255P | 0.802 ms | 0 - 39 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.704 ms | 0 - 38 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.197 ms | 0 - 39 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.328 ms | 0 - 38 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.17 ms | 0 - 43 MB | NPU
| PoseLandmarkDetector | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.321 ms | 0 - 0 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.609 ms | 0 - 80 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 3.204 ms | 0 - 50 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 0.805 ms | 0 - 2 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® SA8775P | 1.431 ms | 0 - 53 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® QCS9075 | 1.231 ms | 0 - 10 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1.169 ms | 0 - 74 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® SA7255P | 3.204 ms | 0 - 50 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Qualcomm® SA8295P | 1.426 ms | 0 - 39 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.499 ms | 0 - 45 MB | NPU
| PoseLandmarkDetector | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.421 ms | 0 - 53 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.213 ms | 0 - 52 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCS6490 | 0.958 ms | 0 - 9 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 0.822 ms | 0 - 36 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.313 ms | 0 - 2 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® SA8775P | 0.548 ms | 0 - 41 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCS9075 | 0.504 ms | 0 - 6 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCM6690 | 1.637 ms | 0 - 34 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.451 ms | 0 - 55 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® SA7255P | 0.822 ms | 0 - 36 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Qualcomm® SA8295P | 0.749 ms | 0 - 36 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.201 ms | 0 - 34 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.326 ms | 0 - 35 MB | NPU
| PoseLandmarkDetector | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.172 ms | 0 - 40 MB | NPU

## License
* The license for the original implementation of MediaPipe-Pose-Estimation can be found
  [here](https://github.com/zmurez/MediaPipePyTorch/blob/master/LICENSE).

## References
* [BlazePose: On-device Real-time Body Pose tracking](https://arxiv.org/abs/2006.10204)
* [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).