v0.30.5
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.30.5 for changelog.
README.md
CHANGED
@@ -37,35 +37,35 @@ More details on model performance across various devices, can be found
|
|
37 |
|
38 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
39 |
|---|---|---|---|---|---|---|---|---|
|
40 |
-
| Unet-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE |
|
41 |
-
| Unet-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN |
|
42 |
-
| Unet-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE |
|
43 |
-
| Unet-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN |
|
44 |
-
| Unet-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE |
|
45 |
-
| Unet-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN |
|
46 |
-
| Unet-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE |
|
47 |
-
| Unet-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN |
|
48 |
-
| Unet-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE |
|
49 |
-
| Unet-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | QNN |
|
50 |
-
| Unet-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE |
|
51 |
-
| Unet-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN |
|
52 |
-
| Unet-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE |
|
53 |
-
| Unet-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | QNN |
|
54 |
-
| Unet-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 161.
|
55 |
-
| Unet-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN |
|
56 |
-
| Unet-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE |
|
57 |
-
| Unet-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | QNN |
|
58 |
-
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE |
|
59 |
-
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN |
|
60 |
-
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX |
|
61 |
-
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE |
|
62 |
-
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN |
|
63 |
-
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX |
|
64 |
-
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE |
|
65 |
-
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN |
|
66 |
-
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX |
|
67 |
-
| Unet-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN |
|
68 |
-
| Unet-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX |
|
69 |
|
70 |
|
71 |
|
@@ -129,8 +129,8 @@ Profiling Results
|
|
129 |
Unet-Segmentation
|
130 |
Device : cs_8275 (ANDROID 14)
|
131 |
Runtime : TFLITE
|
132 |
-
Estimated inference time (ms) :
|
133 |
-
Estimated peak memory usage (MB): [
|
134 |
Total # Ops : 32
|
135 |
Compute Unit(s) : npu (32 ops) gpu (0 ops) cpu (0 ops)
|
136 |
```
|
@@ -219,13 +219,13 @@ AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
|
219 |
You can also run the demo on-device.
|
220 |
|
221 |
```bash
|
222 |
-
python -m qai_hub_models.models.unet_segmentation.demo --on-device
|
223 |
```
|
224 |
|
225 |
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
226 |
environment, please add the following to your cell (instead of the above).
|
227 |
```
|
228 |
-
%run -m qai_hub_models.models.unet_segmentation.demo -- --on-device
|
229 |
```
|
230 |
|
231 |
|
|
|
37 |
|
38 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
39 |
|---|---|---|---|---|---|---|---|---|
|
40 |
+
| Unet-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 7311.581 ms | 2 - 100 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
41 |
+
| Unet-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 7304.677 ms | 5 - 14 MB | NPU | Use Export Script |
|
42 |
+
| Unet-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 262.045 ms | 6 - 161 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
43 |
+
| Unet-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 279.629 ms | 9 - 127 MB | NPU | Use Export Script |
|
44 |
+
| Unet-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 156.435 ms | 6 - 473 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
45 |
+
| Unet-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 145.707 ms | 10 - 12 MB | NPU | Use Export Script |
|
46 |
+
| Unet-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 253.292 ms | 6 - 104 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
47 |
+
| Unet-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 249.863 ms | 0 - 10 MB | NPU | Use Export Script |
|
48 |
+
| Unet-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 7311.581 ms | 2 - 100 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
49 |
+
| Unet-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 7304.677 ms | 5 - 14 MB | NPU | Use Export Script |
|
50 |
+
| Unet-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 157.63 ms | 6 - 464 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
51 |
+
| Unet-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 148.174 ms | 10 - 12 MB | NPU | Use Export Script |
|
52 |
+
| Unet-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 280.587 ms | 4 - 105 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
53 |
+
| Unet-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 276.216 ms | 0 - 17 MB | NPU | Use Export Script |
|
54 |
+
| Unet-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 161.71 ms | 6 - 464 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
55 |
+
| Unet-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 147.53 ms | 10 - 11 MB | NPU | Use Export Script |
|
56 |
+
| Unet-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 253.292 ms | 6 - 104 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
57 |
+
| Unet-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 249.863 ms | 0 - 10 MB | NPU | Use Export Script |
|
58 |
+
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 166.244 ms | 3 - 468 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
59 |
+
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 148.565 ms | 9 - 54 MB | NPU | Use Export Script |
|
60 |
+
| Unet-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 161.453 ms | 13 - 219 MB | NPU | [Unet-Segmentation.onnx](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.onnx) |
|
61 |
+
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 114.828 ms | 5 - 144 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
62 |
+
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 113.855 ms | 9 - 115 MB | NPU | Use Export Script |
|
63 |
+
| Unet-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 118.472 ms | 21 - 124 MB | NPU | [Unet-Segmentation.onnx](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.onnx) |
|
64 |
+
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 107.383 ms | 6 - 106 MB | NPU | [Unet-Segmentation.tflite](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.tflite) |
|
65 |
+
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 94.069 ms | 9 - 126 MB | NPU | Use Export Script |
|
66 |
+
| Unet-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 109.082 ms | 21 - 128 MB | NPU | [Unet-Segmentation.onnx](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.onnx) |
|
67 |
+
| Unet-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 142.445 ms | 9 - 9 MB | NPU | Use Export Script |
|
68 |
+
| Unet-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 151.215 ms | 53 - 53 MB | NPU | [Unet-Segmentation.onnx](https://huggingface.co/qualcomm/Unet-Segmentation/blob/main/Unet-Segmentation.onnx) |
|
69 |
|
70 |
|
71 |
|
|
|
129 |
Unet-Segmentation
|
130 |
Device : cs_8275 (ANDROID 14)
|
131 |
Runtime : TFLITE
|
132 |
+
Estimated inference time (ms) : 7311.6
|
133 |
+
Estimated peak memory usage (MB): [2, 100]
|
134 |
Total # Ops : 32
|
135 |
Compute Unit(s) : npu (32 ops) gpu (0 ops) cpu (0 ops)
|
136 |
```
|
|
|
219 |
You can also run the demo on-device.
|
220 |
|
221 |
```bash
|
222 |
+
python -m qai_hub_models.models.unet_segmentation.demo --eval-mode on-device
|
223 |
```
|
224 |
|
225 |
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
226 |
environment, please add the following to your cell (instead of the above).
|
227 |
```
|
228 |
+
%run -m qai_hub_models.models.unet_segmentation.demo -- --eval-mode on-device
|
229 |
```
|
230 |
|
231 |
|