Hugging Face's logo

qualcomm
/
HRNetPoseQuantized

Keypoint Detection
PyTorch
LiteRT
ONNX
quantized
android
Model card Files Community
qaihm-bot commited on
Commit
524e77c
·
verified ·
1 Parent(s): 54683f3

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +37 -37
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  library_name: pytorch
3
- license: other
4
  tags:
5
  - quantized
6
  - android
@@ -16,7 +16,7 @@ pipeline_tag: keypoint-detection
16
 
17
  HRNet performs pose estimation in high-resolution representations.
18
 
19
- This model is an implementation of HRNetPoseQuantized found [here](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/hrnet_posenet).
20
 
21
 
22
  This repository provides scripts to run HRNetPoseQuantized on Qualcomm® devices.
@@ -35,38 +35,38 @@ More details on model performance across various devices, can be found
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
- | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.988 ms | 0 - 131 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
39
- | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.131 ms | 0 - 3 MB | INT8 | NPU | [HRNetPoseQuantized.so](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.so) |
40
- | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 4.587 ms | 0 - 108 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
41
- | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.715 ms | 0 - 90 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
42
- | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.833 ms | 0 - 18 MB | INT8 | NPU | [HRNetPoseQuantized.so](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.so) |
43
- | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.383 ms | 0 - 163 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
44
- | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.586 ms | 0 - 49 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
45
- | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.693 ms | 0 - 52 MB | INT8 | NPU | Use Export Script |
46
- | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.935 ms | 0 - 120 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
47
- | HRNetPose | SA7255P ADP | SA7255P | TFLITE | 13.956 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
48
- | HRNetPose | SA7255P ADP | SA7255P | QNN | 14.2 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
49
- | HRNetPose | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.98 ms | 0 - 131 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
50
- | HRNetPose | SA8255 (Proxy) | SA8255P Proxy | QNN | 1.14 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
51
- | HRNetPose | SA8295P ADP | SA8295P | TFLITE | 1.714 ms | 0 - 50 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
52
- | HRNetPose | SA8295P ADP | SA8295P | QNN | 1.871 ms | 0 - 18 MB | INT8 | NPU | Use Export Script |
53
- | HRNetPose | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.978 ms | 0 - 134 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
54
- | HRNetPose | SA8650 (Proxy) | SA8650P Proxy | QNN | 1.149 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
55
- | HRNetPose | SA8775P ADP | SA8775P | TFLITE | 1.44 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
56
- | HRNetPose | SA8775P ADP | SA8775P | QNN | 1.585 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
57
- | HRNetPose | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 3.842 ms | 0 - 75 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
58
- | HRNetPose | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 5.281 ms | 0 - 14 MB | INT8 | NPU | Use Export Script |
59
- | HRNetPose | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 17.054 ms | 0 - 2 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
60
- | HRNetPose | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 13.956 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
61
- | HRNetPose | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 14.2 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
62
- | HRNetPose | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.963 ms | 0 - 132 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
63
- | HRNetPose | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 1.134 ms | 0 - 4 MB | INT8 | NPU | Use Export Script |
64
- | HRNetPose | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 1.44 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
65
- | HRNetPose | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 1.585 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
66
- | HRNetPose | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 1.366 ms | 0 - 94 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
67
- | HRNetPose | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.506 ms | 0 - 82 MB | INT8 | NPU | Use Export Script |
68
- | HRNetPose | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.303 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
69
- | HRNetPose | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.826 ms | 26 - 26 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
70
 
71
 
72
 
@@ -131,7 +131,7 @@ HRNetPose
131
  Device : Samsung Galaxy S23 (13)
132
  Runtime : TFLITE
133
  Estimated inference time (ms) : 1.0
134
- Estimated peak memory usage (MB): [0, 131]
135
  Total # Ops : 518
136
  Compute Unit(s) : NPU (518 ops)
137
  ```
@@ -237,14 +237,14 @@ Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
237
 
238
  ## License
239
  * The license for the original implementation of HRNetPoseQuantized can be found
240
- [here](https://github.com/quic/aimet-model-zoo/blob/develop/LICENSE.pdf).
241
  * 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)
242
 
243
 
244
 
245
  ## References
246
  * [Deep High-Resolution Representation Learning for Human Pose Estimation](https://arxiv.org/abs/1902.09212)
247
- * [Source Model Implementation](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/hrnet_posenet)
248
 
249
 
250
 
 
1
  ---
2
  library_name: pytorch
3
+ license: mit
4
  tags:
5
  - quantized
6
  - android
 
16
 
17
  HRNet performs pose estimation in high-resolution representations.
18
 
19
+ This model is an implementation of HRNetPoseQuantized found [here](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch).
20
 
21
 
22
  This repository provides scripts to run HRNetPoseQuantized on Qualcomm® devices.
 
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.984 ms | 0 - 134 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
39
+ | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.178 ms | 0 - 134 MB | INT8 | NPU | [HRNetPoseQuantized.so](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.so) |
40
+ | HRNetPose | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 4.781 ms | 3 - 109 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
41
+ | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.745 ms | 0 - 93 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
42
+ | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.882 ms | 0 - 75 MB | INT8 | NPU | [HRNetPoseQuantized.so](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.so) |
43
+ | HRNetPose | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.41 ms | 5 - 166 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
44
+ | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.692 ms | 0 - 53 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
45
+ | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.774 ms | 0 - 50 MB | INT8 | NPU | Use Export Script |
46
+ | HRNetPose | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.277 ms | 0 - 120 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
47
+ | HRNetPose | SA7255P ADP | SA7255P | TFLITE | 13.949 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
48
+ | HRNetPose | SA7255P ADP | SA7255P | QNN | 14.185 ms | 0 - 9 MB | INT8 | NPU | Use Export Script |
49
+ | HRNetPose | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.987 ms | 0 - 133 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
50
+ | HRNetPose | SA8255 (Proxy) | SA8255P Proxy | QNN | 1.151 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
51
+ | HRNetPose | SA8295P ADP | SA8295P | TFLITE | 1.728 ms | 0 - 50 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
52
+ | HRNetPose | SA8295P ADP | SA8295P | QNN | 1.861 ms | 0 - 18 MB | INT8 | NPU | Use Export Script |
53
+ | HRNetPose | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.988 ms | 0 - 134 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
54
+ | HRNetPose | SA8650 (Proxy) | SA8650P Proxy | QNN | 1.135 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
55
+ | HRNetPose | SA8775P ADP | SA8775P | TFLITE | 1.451 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
56
+ | HRNetPose | SA8775P ADP | SA8775P | QNN | 1.595 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
57
+ | HRNetPose | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 3.792 ms | 0 - 74 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
58
+ | HRNetPose | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 5.279 ms | 0 - 11 MB | INT8 | NPU | Use Export Script |
59
+ | HRNetPose | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 17.161 ms | 0 - 2 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
60
+ | HRNetPose | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 13.949 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
61
+ | HRNetPose | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 14.185 ms | 0 - 9 MB | INT8 | NPU | Use Export Script |
62
+ | HRNetPose | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.991 ms | 0 - 134 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
63
+ | HRNetPose | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 1.142 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
64
+ | HRNetPose | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 1.451 ms | 0 - 48 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
65
+ | HRNetPose | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 1.595 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
66
+ | HRNetPose | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 1.301 ms | 0 - 92 MB | INT8 | NPU | [HRNetPoseQuantized.tflite](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.tflite) |
67
+ | HRNetPose | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.587 ms | 0 - 79 MB | INT8 | NPU | Use Export Script |
68
+ | HRNetPose | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.267 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
69
+ | HRNetPose | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.892 ms | 27 - 27 MB | INT8 | NPU | [HRNetPoseQuantized.onnx](https://huggingface.co/qualcomm/HRNetPoseQuantized/blob/main/HRNetPose.onnx) |
70
 
71
 
72
 
 
131
  Device : Samsung Galaxy S23 (13)
132
  Runtime : TFLITE
133
  Estimated inference time (ms) : 1.0
134
+ Estimated peak memory usage (MB): [0, 134]
135
  Total # Ops : 518
136
  Compute Unit(s) : NPU (518 ops)
137
  ```
 
237
 
238
  ## License
239
  * The license for the original implementation of HRNetPoseQuantized can be found
240
+ [here](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch/blob/master/LICENSE).
241
  * 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)
242
 
243
 
244
 
245
  ## References
246
  * [Deep High-Resolution Representation Learning for Human Pose Estimation](https://arxiv.org/abs/1902.09212)
247
+ * [Source Model Implementation](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch)
248
 
249
 
250