qaihm-bot commited on
Commit
a3151a5
·
verified ·
1 Parent(s): cde5d07

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +55 -38
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  library_name: pytorch
3
- license: mit
4
  tags:
5
  - real_time
6
  - android
@@ -26,7 +26,7 @@ More details on model performance across various devices, can be found
26
 
27
  ### Model Details
28
 
29
- - **Model Type:** Semantic segmentation
30
  - **Model Stats:**
31
  - Model checkpoint: PIDNet_S_Cityscapes_val.pt
32
  - Inference latency: RealTime
@@ -35,37 +35,54 @@ More details on model performance across various devices, can be found
35
  - Number of parameters: 7.62M
36
  - Model size: 29.1 MB
37
 
38
- | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
39
  |---|---|---|---|---|---|---|---|---|
40
- | PidNet | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 46.93 ms | 2 - 20 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
41
- | PidNet | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 47.316 ms | 24 - 40 MB | FP16 | NPU | [PidNet.so](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.so) |
42
- | PidNet | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 37.009 ms | 29 - 79 MB | FP16 | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
43
- | PidNet | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 31.696 ms | 1 - 64 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
44
- | PidNet | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 31.624 ms | 24 - 69 MB | FP16 | NPU | [PidNet.so](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.so) |
45
- | PidNet | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 25.751 ms | 30 - 91 MB | FP16 | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
46
- | PidNet | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 26.311 ms | 1 - 53 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
47
- | PidNet | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 32.027 ms | 22 - 67 MB | FP16 | NPU | Use Export Script |
48
- | PidNet | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 20.977 ms | 27 - 82 MB | FP16 | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
49
- | PidNet | SA7255P ADP | SA7255P | TFLITE | 687.533 ms | 0 - 49 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
50
- | PidNet | SA7255P ADP | SA7255P | QNN | 663.511 ms | 24 - 34 MB | FP16 | NPU | Use Export Script |
51
- | PidNet | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 46.982 ms | 2 - 19 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
52
- | PidNet | SA8255 (Proxy) | SA8255P Proxy | QNN | 36.224 ms | 4 - 6 MB | FP16 | NPU | Use Export Script |
53
- | PidNet | SA8295P ADP | SA8295P | TFLITE | 66.483 ms | 2 - 50 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
54
- | PidNet | SA8295P ADP | SA8295P | QNN | 53.27 ms | 24 - 42 MB | FP16 | NPU | Use Export Script |
55
- | PidNet | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 46.856 ms | 2 - 23 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
56
- | PidNet | SA8650 (Proxy) | SA8650P Proxy | QNN | 35.91 ms | 25 - 28 MB | FP16 | NPU | Use Export Script |
57
- | PidNet | SA8775P ADP | SA8775P | TFLITE | 70.226 ms | 2 - 52 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
58
- | PidNet | SA8775P ADP | SA8775P | QNN | 54.086 ms | 24 - 34 MB | FP16 | NPU | Use Export Script |
59
- | PidNet | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 687.533 ms | 0 - 49 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
60
- | PidNet | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 663.511 ms | 24 - 34 MB | FP16 | NPU | Use Export Script |
61
- | PidNet | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 46.906 ms | 2 - 29 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
62
- | PidNet | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 35.571 ms | 24 - 26 MB | FP16 | NPU | Use Export Script |
63
- | PidNet | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 70.226 ms | 2 - 52 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
64
- | PidNet | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 54.086 ms | 24 - 34 MB | FP16 | NPU | Use Export Script |
65
- | PidNet | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 58.795 ms | 2 - 62 MB | FP16 | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
66
- | PidNet | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 58.377 ms | 24 - 69 MB | FP16 | NPU | Use Export Script |
67
- | PidNet | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 36.362 ms | 24 - 24 MB | FP16 | NPU | Use Export Script |
68
- | PidNet | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 40.53 ms | 24 - 24 MB | FP16 | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
 
@@ -127,12 +144,12 @@ python -m qai_hub_models.models.pidnet.export
127
  Profiling Results
128
  ------------------------------------------------------------
129
  PidNet
130
- Device : Samsung Galaxy S23 (13)
131
- Runtime : TFLITE
132
- Estimated inference time (ms) : 46.9
133
- Estimated peak memory usage (MB): [2, 20]
134
- Total # Ops : 169
135
- Compute Unit(s) : NPU (169 ops)
136
  ```
137
 
138
 
 
1
  ---
2
  library_name: pytorch
3
+ license: other
4
  tags:
5
  - real_time
6
  - android
 
26
 
27
  ### Model Details
28
 
29
+ - **Model Type:** Model_use_case.semantic_segmentation
30
  - **Model Stats:**
31
  - Model checkpoint: PIDNet_S_Cityscapes_val.pt
32
  - Inference latency: RealTime
 
35
  - Number of parameters: 7.62M
36
  - Model size: 29.1 MB
37
 
38
+ | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
39
  |---|---|---|---|---|---|---|---|---|
40
+ | PidNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 136.982 ms | 2 - 52 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
41
+ | PidNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 114.542 ms | 24 - 34 MB | NPU | Use Export Script |
42
+ | PidNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 58.674 ms | 2 - 64 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
43
+ | PidNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 63.016 ms | 19 - 94 MB | NPU | Use Export Script |
44
+ | PidNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 46.945 ms | 2 - 21 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
45
+ | PidNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 35.591 ms | 24 - 28 MB | NPU | Use Export Script |
46
+ | PidNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 57.111 ms | 0 - 50 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
47
+ | PidNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 44.65 ms | 22 - 37 MB | NPU | Use Export Script |
48
+ | PidNet | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 136.982 ms | 2 - 52 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
49
+ | PidNet | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 114.542 ms | 24 - 34 MB | NPU | Use Export Script |
50
+ | PidNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 47.006 ms | 2 - 23 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
51
+ | PidNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 35.616 ms | 24 - 26 MB | NPU | Use Export Script |
52
+ | PidNet | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 66.478 ms | 2 - 51 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
53
+ | PidNet | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 53.518 ms | 24 - 42 MB | NPU | Use Export Script |
54
+ | PidNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 46.99 ms | 2 - 19 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
55
+ | PidNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 35.818 ms | 26 - 28 MB | NPU | Use Export Script |
56
+ | PidNet | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 57.111 ms | 0 - 50 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
57
+ | PidNet | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 44.65 ms | 22 - 37 MB | NPU | Use Export Script |
58
+ | PidNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 46.956 ms | 2 - 19 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
59
+ | PidNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 36.135 ms | 22 - 51 MB | NPU | Use Export Script |
60
+ | PidNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 37.55 ms | 29 - 87 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
61
+ | PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 31.732 ms | 1 - 67 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
62
+ | PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 25.665 ms | 24 - 91 MB | NPU | Use Export Script |
63
+ | PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 26.068 ms | 28 - 89 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
64
+ | PidNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 31.988 ms | 2 - 54 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.tflite) |
65
+ | PidNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 23.019 ms | 16 - 84 MB | NPU | Use Export Script |
66
+ | PidNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 18.209 ms | 23 - 78 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
67
+ | PidNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 36.402 ms | 24 - 24 MB | NPU | Use Export Script |
68
+ | PidNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 40.641 ms | 24 - 24 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet.onnx) |
69
+ | PidNet | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 100.964 ms | 1 - 37 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
70
+ | PidNet | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 51.18 ms | 1 - 53 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
71
+ | PidNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 50.103 ms | 1 - 30 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
72
+ | PidNet | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 51.097 ms | 1 - 38 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
73
+ | PidNet | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 165.666 ms | 1 - 41 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
74
+ | PidNet | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 100.964 ms | 1 - 37 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
75
+ | PidNet | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 50.348 ms | 1 - 30 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
76
+ | PidNet | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 58.722 ms | 1 - 39 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
77
+ | PidNet | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 50.215 ms | 1 - 24 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
78
+ | PidNet | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 51.097 ms | 1 - 38 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
79
+ | PidNet | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 50.145 ms | 1 - 22 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
80
+ | PidNet | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 80.569 ms | 68 - 102 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.onnx) |
81
+ | PidNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 38.486 ms | 1 - 49 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
82
+ | PidNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 65.097 ms | 85 - 214 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.onnx) |
83
+ | PidNet | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 47.058 ms | 1 - 43 MB | NPU | [PidNet.tflite](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.tflite) |
84
+ | PidNet | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 56.149 ms | 59 - 181 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.onnx) |
85
+ | PidNet | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 107.732 ms | 134 - 134 MB | NPU | [PidNet.onnx](https://huggingface.co/qualcomm/PidNet/blob/main/PidNet_w8a8.onnx) |
86
 
87
 
88
 
 
144
  Profiling Results
145
  ------------------------------------------------------------
146
  PidNet
147
+ Device : cs_8275 (ANDROID 14)
148
+ Runtime : TFLITE
149
+ Estimated inference time (ms) : 137.0
150
+ Estimated peak memory usage (MB): [2, 52]
151
+ Total # Ops : 169
152
+ Compute Unit(s) : npu (169 ops) gpu (0 ops) cpu (0 ops)
153
  ```
154
 
155