v0.42.0
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.42.0 for changelog.
README.md
CHANGED
|
@@ -38,42 +38,44 @@ More details on model performance across various devices, can be found
|
|
| 38 |
|
| 39 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 40 |
|---|---|---|---|---|---|---|---|---|
|
| 41 |
-
| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 64.
|
| 42 |
-
| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC |
|
| 43 |
-
| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 23.
|
| 44 |
-
| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC |
|
| 45 |
-
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 11.
|
| 46 |
-
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 10.
|
| 47 |
-
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 13.
|
| 48 |
-
| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 80.
|
| 49 |
-
| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.
|
| 50 |
-
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 8.
|
| 51 |
-
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 8.
|
| 52 |
-
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 9.
|
| 53 |
-
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 6.
|
| 54 |
-
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 6.
|
| 55 |
-
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 7.
|
| 56 |
-
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 4.
|
| 57 |
-
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 4.
|
| 58 |
-
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 5.
|
| 59 |
-
| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 11.
|
| 60 |
-
| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 13.
|
| 61 |
-
| Yolo-v5 | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC |
|
| 62 |
-
| Yolo-v5 | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC |
|
| 63 |
-
| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 8.
|
| 64 |
-
| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX |
|
| 65 |
-
| Yolo-v5 | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC |
|
| 66 |
-
| Yolo-v5 | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 48.
|
| 67 |
-
| Yolo-v5 | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX |
|
| 68 |
-
| Yolo-v5 | w8a16 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX |
|
| 69 |
-
| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 5.
|
| 70 |
-
| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 8.
|
| 71 |
-
| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 4.25 ms |
|
| 72 |
-
| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 5.
|
| 73 |
-
| Yolo-v5 | w8a16 | Snapdragon
|
| 74 |
-
| Yolo-v5 | w8a16 | Snapdragon
|
| 75 |
-
| Yolo-v5 | w8a16 | Snapdragon
|
| 76 |
-
| Yolo-v5 | w8a16 | Snapdragon
|
|
|
|
|
|
|
| 77 |
|
| 78 |
|
| 79 |
|
|
@@ -88,9 +90,9 @@ pip install "qai-hub-models[yolov5]"
|
|
| 88 |
```
|
| 89 |
|
| 90 |
|
| 91 |
-
## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
|
| 92 |
|
| 93 |
-
Sign-in to [Qualcomm® AI Hub](https://
|
| 94 |
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
|
| 95 |
|
| 96 |
With this API token, you can configure your client to run models on the cloud
|
|
@@ -98,7 +100,7 @@ hosted devices.
|
|
| 98 |
```bash
|
| 99 |
qai-hub configure --api_token API_TOKEN
|
| 100 |
```
|
| 101 |
-
Navigate to [docs](https://
|
| 102 |
|
| 103 |
|
| 104 |
|
|
@@ -209,7 +211,7 @@ With the output of the model, you can compute like PSNR, relative errors or
|
|
| 209 |
spot check the output with expected output.
|
| 210 |
|
| 211 |
**Note**: This on-device profiling and inference requires access to Qualcomm®
|
| 212 |
-
AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
| 213 |
|
| 214 |
|
| 215 |
|
|
|
|
| 38 |
|
| 39 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 40 |
|---|---|---|---|---|---|---|---|---|
|
| 41 |
+
| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 64.108 ms | 0 - 80 MB | NPU | -- |
|
| 42 |
+
| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 63.282 ms | 0 - 90 MB | NPU | -- |
|
| 43 |
+
| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 23.595 ms | 0 - 97 MB | NPU | -- |
|
| 44 |
+
| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 25.517 ms | 5 - 58 MB | NPU | -- |
|
| 45 |
+
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 11.654 ms | 0 - 20 MB | NPU | -- |
|
| 46 |
+
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 10.877 ms | 5 - 84 MB | NPU | -- |
|
| 47 |
+
| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 13.081 ms | 0 - 79 MB | NPU | -- |
|
| 48 |
+
| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 80.325 ms | 0 - 79 MB | NPU | -- |
|
| 49 |
+
| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.048 ms | 2 - 89 MB | NPU | -- |
|
| 50 |
+
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 8.667 ms | 0 - 113 MB | NPU | -- |
|
| 51 |
+
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 8.244 ms | 5 - 163 MB | NPU | -- |
|
| 52 |
+
| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 9.639 ms | 5 - 85 MB | NPU | -- |
|
| 53 |
+
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 6.938 ms | 0 - 82 MB | NPU | -- |
|
| 54 |
+
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 6.163 ms | 5 - 98 MB | NPU | -- |
|
| 55 |
+
| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 7.541 ms | 2 - 84 MB | NPU | -- |
|
| 56 |
+
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 4.952 ms | 0 - 79 MB | NPU | -- |
|
| 57 |
+
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 4.503 ms | 5 - 110 MB | NPU | -- |
|
| 58 |
+
| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 5.928 ms | 5 - 72 MB | NPU | -- |
|
| 59 |
+
| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 11.532 ms | 102 - 102 MB | NPU | -- |
|
| 60 |
+
| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 13.332 ms | 46 - 46 MB | NPU | -- |
|
| 61 |
+
| Yolo-v5 | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 20.837 ms | 2 - 65 MB | NPU | -- |
|
| 62 |
+
| Yolo-v5 | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 12.526 ms | 2 - 83 MB | NPU | -- |
|
| 63 |
+
| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 8.633 ms | 2 - 19 MB | NPU | -- |
|
| 64 |
+
| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 11.959 ms | 0 - 75 MB | NPU | -- |
|
| 65 |
+
| Yolo-v5 | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 8.836 ms | 1 - 64 MB | NPU | -- |
|
| 66 |
+
| Yolo-v5 | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 48.604 ms | 2 - 101 MB | NPU | -- |
|
| 67 |
+
| Yolo-v5 | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 814.564 ms | 171 - 189 MB | CPU | -- |
|
| 68 |
+
| Yolo-v5 | w8a16 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 794.538 ms | 158 - 170 MB | CPU | -- |
|
| 69 |
+
| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 5.74 ms | 2 - 81 MB | NPU | -- |
|
| 70 |
+
| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 8.655 ms | 2 - 128 MB | NPU | -- |
|
| 71 |
+
| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 4.25 ms | 2 - 74 MB | NPU | -- |
|
| 72 |
+
| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 5.25 ms | 0 - 101 MB | NPU | -- |
|
| 73 |
+
| Yolo-v5 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 12.367 ms | 2 - 82 MB | NPU | -- |
|
| 74 |
+
| Yolo-v5 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 819.861 ms | 160 - 178 MB | CPU | -- |
|
| 75 |
+
| Yolo-v5 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 3.25 ms | 2 - 73 MB | NPU | -- |
|
| 76 |
+
| Yolo-v5 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 3.983 ms | 2 - 95 MB | NPU | -- |
|
| 77 |
+
| Yolo-v5 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 9.328 ms | 64 - 64 MB | NPU | -- |
|
| 78 |
+
| Yolo-v5 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 11.945 ms | 24 - 24 MB | NPU | -- |
|
| 79 |
|
| 80 |
|
| 81 |
|
|
|
|
| 90 |
```
|
| 91 |
|
| 92 |
|
| 93 |
+
## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
|
| 94 |
|
| 95 |
+
Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
|
| 96 |
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
|
| 97 |
|
| 98 |
With this API token, you can configure your client to run models on the cloud
|
|
|
|
| 100 |
```bash
|
| 101 |
qai-hub configure --api_token API_TOKEN
|
| 102 |
```
|
| 103 |
+
Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
|
| 104 |
|
| 105 |
|
| 106 |
|
|
|
|
| 211 |
spot check the output with expected output.
|
| 212 |
|
| 213 |
**Note**: This on-device profiling and inference requires access to Qualcomm®
|
| 214 |
+
AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
| 215 |
|
| 216 |
|
| 217 |
|