Hugging Face's logo

qualcomm
/
FCN-ResNet50-Quantized

Image Segmentation
PyTorch
TF Lite
ONNX
quantized
android
Model card Files Community
bhushans commited on
Commit
960a093
·
verified ·
1 Parent(s): 4cbcb7a

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +42 -24
README.md CHANGED
@@ -36,27 +36,29 @@ More details on model performance across various devices, can be found
36
 
37
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
- | FCN-ResNet50-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 15.313 ms | 0 - 2 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
40
- | FCN-ResNet50-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 17.154 ms | 0 - 15 MB | INT8 | NPU | [FCN-ResNet50-Quantized.so](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.so) |
41
- | FCN-ResNet50-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 11.131 ms | 0 - 106 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
42
- | FCN-ResNet50-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 12.736 ms | 1 - 43 MB | INT8 | NPU | [FCN-ResNet50-Quantized.so](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.so) |
43
- | FCN-ResNet50-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 9.335 ms | 0 - 55 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
44
- | FCN-ResNet50-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 12.821 ms | 1 - 40 MB | INT8 | NPU | Use Export Script |
45
- | FCN-ResNet50-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 140.471 ms | 2 - 10 MB | INT8 | NPU | Use Export Script |
46
- | FCN-ResNet50-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 1390.981 ms | 86 - 158 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
47
- | FCN-ResNet50-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 15.309 ms | 0 - 330 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
48
- | FCN-ResNet50-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 14.137 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
49
- | FCN-ResNet50-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 15.238 ms | 0 - 2 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
50
- | FCN-ResNet50-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 14.236 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
51
- | FCN-ResNet50-Quantized | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 15.292 ms | 0 - 2 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
52
- | FCN-ResNet50-Quantized | SA8775 (Proxy) | SA8775P Proxy | QNN | 14.211 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
53
- | FCN-ResNet50-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 15.275 ms | 0 - 2 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
54
- | FCN-ResNet50-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 14.194 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
55
- | FCN-ResNet50-Quantized | SA8295P ADP | SA8295P | TFLITE | 22.596 ms | 0 - 53 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
56
- | FCN-ResNet50-Quantized | SA8295P ADP | SA8295P | QNN | 21.575 ms | 1 - 6 MB | INT8 | NPU | Use Export Script |
57
- | FCN-ResNet50-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 17.951 ms | 1 - 106 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
58
- | FCN-ResNet50-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 19.842 ms | 1 - 42 MB | INT8 | NPU | Use Export Script |
59
- | FCN-ResNet50-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 14.608 ms | 1 - 1 MB | INT8 | NPU | Use Export Script |
 
 
60
 
61
 
62
 
@@ -121,8 +123,8 @@ Profiling Results
121
  FCN-ResNet50-Quantized
122
  Device : Samsung Galaxy S23 (13)
123
  Runtime : TFLITE
124
- Estimated inference time (ms) : 15.3
125
- Estimated peak memory usage (MB): [0, 2]
126
  Total # Ops : 90
127
  Compute Unit(s) : NPU (90 ops)
128
  ```
@@ -143,13 +145,29 @@ in memory using the `jit.trace` and then call the `submit_compile_job` API.
143
  import torch
144
 
145
  import qai_hub as hub
146
- from qai_hub_models.models.fcn_resnet50_quantized import
147
 
148
  # Load the model
 
149
 
150
  # Device
151
  device = hub.Device("Samsung Galaxy S23")
152
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
153
 
154
  ```
155
 
 
36
 
37
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
+ | FCN-ResNet50-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 15.139 ms | 0 - 18 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
40
+ | FCN-ResNet50-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 17.116 ms | 0 - 18 MB | INT8 | NPU | [FCN-ResNet50-Quantized.so](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.so) |
41
+ | FCN-ResNet50-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 11.353 ms | 0 - 45 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
42
+ | FCN-ResNet50-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 12.899 ms | 1 - 42 MB | INT8 | NPU | [FCN-ResNet50-Quantized.so](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.so) |
43
+ | FCN-ResNet50-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 9.425 ms | 0 - 40 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
44
+ | FCN-ResNet50-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 10.79 ms | 1 - 41 MB | INT8 | NPU | Use Export Script |
45
+ | FCN-ResNet50-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 139.127 ms | 1 - 9 MB | INT8 | NPU | Use Export Script |
46
+ | FCN-ResNet50-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 1383.08 ms | 87 - 158 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
47
+ | FCN-ResNet50-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 15.186 ms | 0 - 16 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
48
+ | FCN-ResNet50-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 14.114 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
49
+ | FCN-ResNet50-Quantized | SA7255P ADP | SA7255P | TFLITE | 270.259 ms | 0 - 39 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
50
+ | FCN-ResNet50-Quantized | SA7255P ADP | SA7255P | QNN | 268.117 ms | 1 - 7 MB | INT8 | NPU | Use Export Script |
51
+ | FCN-ResNet50-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 15.144 ms | 0 - 18 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
52
+ | FCN-ResNet50-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 14.237 ms | 1 - 3 MB | INT8 | NPU | Use Export Script |
53
+ | FCN-ResNet50-Quantized | SA8295P ADP | SA8295P | TFLITE | 22.346 ms | 0 - 39 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
54
+ | FCN-ResNet50-Quantized | SA8295P ADP | SA8295P | QNN | 21.046 ms | 1 - 6 MB | INT8 | NPU | Use Export Script |
55
+ | FCN-ResNet50-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 15.187 ms | 0 - 24 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
56
+ | FCN-ResNet50-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 14.208 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
57
+ | FCN-ResNet50-Quantized | SA8775P ADP | SA8775P | TFLITE | 18.77 ms | 0 - 39 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
58
+ | FCN-ResNet50-Quantized | SA8775P ADP | SA8775P | QNN | 17.772 ms | 1 - 7 MB | INT8 | NPU | Use Export Script |
59
+ | FCN-ResNet50-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 17.903 ms | 0 - 49 MB | INT8 | NPU | [FCN-ResNet50-Quantized.tflite](https://huggingface.co/qualcomm/FCN-ResNet50-Quantized/blob/main/FCN-ResNet50-Quantized.tflite) |
60
+ | FCN-ResNet50-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 19.761 ms | 1 - 44 MB | INT8 | NPU | Use Export Script |
61
+ | FCN-ResNet50-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 14.577 ms | 1 - 1 MB | INT8 | NPU | Use Export Script |
62
 
63
 
64
 
 
123
  FCN-ResNet50-Quantized
124
  Device : Samsung Galaxy S23 (13)
125
  Runtime : TFLITE
126
+ Estimated inference time (ms) : 15.1
127
+ Estimated peak memory usage (MB): [0, 18]
128
  Total # Ops : 90
129
  Compute Unit(s) : NPU (90 ops)
130
  ```
 
145
  import torch
146
 
147
  import qai_hub as hub
148
+ from qai_hub_models.models.fcn_resnet50_quantized import Model
149
 
150
  # Load the model
151
+ torch_model = Model.from_pretrained()
152
 
153
  # Device
154
  device = hub.Device("Samsung Galaxy S23")
155
 
156
+ # Trace model
157
+ input_shape = torch_model.get_input_spec()
158
+ sample_inputs = torch_model.sample_inputs()
159
+
160
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
161
+
162
+ # Compile model on a specific device
163
+ compile_job = hub.submit_compile_job(
164
+ model=pt_model,
165
+ device=device,
166
+ input_specs=torch_model.get_input_spec(),
167
+ )
168
+
169
+ # Get target model to run on-device
170
+ target_model = compile_job.get_target_model()
171
 
172
  ```
173