v0.46.1
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.46.1 for changelog.
- Detectron2-Detection_Detectron2ProposalGenerator_float.dlc +0 -3
- Detectron2-Detection_Detectron2ProposalGenerator_float.tflite +0 -3
- Detectron2-Detection_Detectron2ProposalGenerator_w8a16.dlc +0 -3
- Detectron2-Detection_Detectron2ProposalGenerator_w8a16.onnx.zip +0 -3
- Detectron2-Detection_Detectron2ProposalGenerator_w8a8.dlc +0 -3
- Detectron2-Detection_Detectron2ProposalGenerator_w8a8.onnx.zip +0 -3
- Detectron2-Detection_Detectron2ROIHead_float.dlc +0 -3
- Detectron2-Detection_Detectron2ROIHead_float.tflite +0 -3
- Detectron2-Detection_Detectron2ROIHead_w8a16.dlc +0 -3
- Detectron2-Detection_Detectron2ROIHead_w8a16.onnx.zip +0 -3
- Detectron2-Detection_Detectron2ROIHead_w8a8.dlc +0 -3
- Detectron2-Detection_Detectron2ROIHead_w8a8.onnx.zip +0 -3
- README.md +186 -316
- tool-versions.yaml +0 -4
Detectron2-Detection_Detectron2ProposalGenerator_float.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:731b50dee0bad4bbd56d7cbd3584634a8c976e0a864919b90a60275e83b5fe11
|
| 3 |
-
size 72910388
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ProposalGenerator_float.tflite
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:9940797d7ffdb96332fbd1f50a9af650b40383a0dd539b469b0b0af064b27b43
|
| 3 |
-
size 72802024
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ProposalGenerator_w8a16.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8b06afd371f1f60ca4c6e68a8bfba0fff2a1c46a283af3e043abddc8e31e7833
|
| 3 |
-
size 19455796
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ProposalGenerator_w8a16.onnx.zip
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8abe1e76827b33465edf977eb1178bee434c792f947088b008caee8cebc95f1c
|
| 3 |
-
size 31298196
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ProposalGenerator_w8a8.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:40216fd9ade37cf232737f4258377cf4620fa54f4f23eade6d0e6bacedf3c365
|
| 3 |
-
size 19305788
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ProposalGenerator_w8a8.onnx.zip
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e775ec5f921779872539096d1b97e658f23a3ec19876d73903c0f64f4cb52b3d
|
| 3 |
-
size 31380433
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_float.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:132954b3f5840a0407da8f6d17a4c7e8c8473dc3d8f5f8f9e6b39233358de116
|
| 3 |
-
size 63197524
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_float.tflite
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:fd9cb74ca9a03874b6e486030009a84d7963d2379af01110db24a26d45706e6c
|
| 3 |
-
size 63293948
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_w8a16.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:2344d092b1b7273d3ea604a612d9b966c0e03ca6d68c85e8966ab6cb4d63ff94
|
| 3 |
-
size 16544580
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_w8a16.onnx.zip
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:c7adbff4df7b207d285a6cf73a73ac7baad8fbadcdecd322d2a9a2465d5b1a0b
|
| 3 |
-
size 26485841
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_w8a8.dlc
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:288b074d3cc30eae192b64284c08ccf2039dd1084e1224ce4d6d632c97bb6677
|
| 3 |
-
size 16544372
|
|
|
|
|
|
|
|
|
|
|
|
Detectron2-Detection_Detectron2ROIHead_w8a8.onnx.zip
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:37806372ee10f8e207ed9fa56bc0f11fa0de1d4fa98109abf22d061e957df92a
|
| 3 |
-
size 26587837
|
|
|
|
|
|
|
|
|
|
|
|
README.md
CHANGED
|
@@ -9,333 +9,203 @@ pipeline_tag: object-detection
|
|
| 9 |
|
| 10 |
|
| 11 |
|
| 12 |
-
# Detectron2-Detection: Optimized for
|
| 13 |
-
## A next-generation library for object detection
|
| 14 |
-
|
| 15 |
|
| 16 |
Detectron2-Detection is a machine learning model that can detect objects (trained on COCO dataset).
|
| 17 |
|
| 18 |
-
This
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
|
| 37 |
-
|
|
| 38 |
-
|
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
|
| 66 |
-
|
|
| 67 |
-
| Detectron2ProposalGenerator |
|
| 68 |
-
| Detectron2ProposalGenerator |
|
| 69 |
-
| Detectron2ProposalGenerator |
|
| 70 |
-
| Detectron2ProposalGenerator |
|
| 71 |
-
| Detectron2ProposalGenerator |
|
| 72 |
-
| Detectron2ProposalGenerator |
|
| 73 |
-
| Detectron2ProposalGenerator |
|
| 74 |
-
| Detectron2ProposalGenerator |
|
| 75 |
-
| Detectron2ProposalGenerator |
|
| 76 |
-
| Detectron2ProposalGenerator |
|
| 77 |
-
| Detectron2ProposalGenerator |
|
| 78 |
-
| Detectron2ProposalGenerator |
|
| 79 |
-
| Detectron2ProposalGenerator |
|
| 80 |
-
| Detectron2ProposalGenerator |
|
| 81 |
-
| Detectron2ProposalGenerator |
|
| 82 |
-
| Detectron2ProposalGenerator |
|
| 83 |
-
| Detectron2ProposalGenerator |
|
| 84 |
-
| Detectron2ProposalGenerator |
|
| 85 |
-
| Detectron2ProposalGenerator |
|
| 86 |
-
| Detectron2ProposalGenerator |
|
| 87 |
-
|
|
| 88 |
-
|
|
| 89 |
-
|
|
| 90 |
-
|
|
| 91 |
-
|
|
| 92 |
-
|
|
| 93 |
-
|
|
| 94 |
-
|
|
| 95 |
-
|
|
| 96 |
-
|
|
| 97 |
-
|
|
| 98 |
-
|
|
| 99 |
-
|
|
| 100 |
-
|
|
| 101 |
-
|
|
| 102 |
-
|
|
| 103 |
-
|
|
| 104 |
-
|
|
| 105 |
-
|
|
| 106 |
-
|
|
| 107 |
-
|
|
| 108 |
-
| Detectron2ProposalGenerator |
|
| 109 |
-
| Detectron2ProposalGenerator |
|
| 110 |
-
| Detectron2ProposalGenerator |
|
| 111 |
-
| Detectron2ProposalGenerator |
|
| 112 |
-
| Detectron2ProposalGenerator |
|
| 113 |
-
| Detectron2ProposalGenerator |
|
| 114 |
-
| Detectron2ProposalGenerator |
|
| 115 |
-
| Detectron2ProposalGenerator |
|
| 116 |
-
| Detectron2ProposalGenerator |
|
| 117 |
-
| Detectron2ProposalGenerator |
|
| 118 |
-
| Detectron2ProposalGenerator |
|
| 119 |
-
| Detectron2ProposalGenerator |
|
| 120 |
-
| Detectron2ProposalGenerator |
|
| 121 |
-
| Detectron2ProposalGenerator |
|
| 122 |
-
| Detectron2ProposalGenerator |
|
| 123 |
-
| Detectron2ProposalGenerator |
|
| 124 |
-
| Detectron2ProposalGenerator |
|
| 125 |
-
| Detectron2ProposalGenerator |
|
| 126 |
-
| Detectron2ProposalGenerator |
|
| 127 |
-
| Detectron2ProposalGenerator |
|
| 128 |
-
| Detectron2ProposalGenerator |
|
| 129 |
-
| Detectron2ProposalGenerator |
|
| 130 |
-
|
|
| 131 |
-
|
|
| 132 |
-
|
|
| 133 |
-
|
|
| 134 |
-
|
|
| 135 |
-
| Detectron2ROIHead |
|
| 136 |
-
| Detectron2ROIHead |
|
| 137 |
-
| Detectron2ROIHead |
|
| 138 |
-
| Detectron2ROIHead |
|
| 139 |
-
| Detectron2ROIHead |
|
| 140 |
-
| Detectron2ROIHead |
|
| 141 |
-
| Detectron2ROIHead |
|
| 142 |
-
| Detectron2ROIHead |
|
| 143 |
-
| Detectron2ROIHead |
|
| 144 |
-
| Detectron2ROIHead |
|
| 145 |
-
| Detectron2ROIHead |
|
| 146 |
-
| Detectron2ROIHead |
|
| 147 |
-
| Detectron2ROIHead |
|
| 148 |
-
| Detectron2ROIHead |
|
| 149 |
-
| Detectron2ROIHead |
|
| 150 |
-
| Detectron2ROIHead |
|
| 151 |
-
| Detectron2ROIHead |
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
Qualcomm®
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
device. This script does the following:
|
| 204 |
-
* Performance check on-device on a cloud-hosted device
|
| 205 |
-
* Downloads compiled assets that can be deployed on-device for Android.
|
| 206 |
-
* Accuracy check between PyTorch and on-device outputs.
|
| 207 |
-
|
| 208 |
-
```bash
|
| 209 |
-
python -m qai_hub_models.models.detectron2_detection.export
|
| 210 |
-
```
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
## How does this work?
|
| 215 |
-
|
| 216 |
-
This [export script](https://aihub.qualcomm.com/models/detectron2_detection/qai_hub_models/models/Detectron2-Detection/export.py)
|
| 217 |
-
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
|
| 218 |
-
on-device. Lets go through each step below in detail:
|
| 219 |
-
|
| 220 |
-
Step 1: **Compile model for on-device deployment**
|
| 221 |
-
|
| 222 |
-
To compile a PyTorch model for on-device deployment, we first trace the model
|
| 223 |
-
in memory using the `jit.trace` and then call the `submit_compile_job` API.
|
| 224 |
-
|
| 225 |
-
```python
|
| 226 |
-
import torch
|
| 227 |
-
|
| 228 |
-
import qai_hub as hub
|
| 229 |
-
from qai_hub_models.models.detectron2_detection import Model
|
| 230 |
-
|
| 231 |
-
# Load the model
|
| 232 |
-
torch_model = Model.from_pretrained()
|
| 233 |
-
|
| 234 |
-
# Device
|
| 235 |
-
device = hub.Device("Samsung Galaxy S25")
|
| 236 |
-
|
| 237 |
-
# Trace model
|
| 238 |
-
input_shape = torch_model.get_input_spec()
|
| 239 |
-
sample_inputs = torch_model.sample_inputs()
|
| 240 |
-
|
| 241 |
-
pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
|
| 242 |
-
|
| 243 |
-
# Compile model on a specific device
|
| 244 |
-
compile_job = hub.submit_compile_job(
|
| 245 |
-
model=pt_model,
|
| 246 |
-
device=device,
|
| 247 |
-
input_specs=torch_model.get_input_spec(),
|
| 248 |
-
)
|
| 249 |
-
|
| 250 |
-
# Get target model to run on-device
|
| 251 |
-
target_model = compile_job.get_target_model()
|
| 252 |
-
|
| 253 |
-
```
|
| 254 |
-
|
| 255 |
-
|
| 256 |
-
Step 2: **Performance profiling on cloud-hosted device**
|
| 257 |
-
|
| 258 |
-
After compiling models from step 1. Models can be profiled model on-device using the
|
| 259 |
-
`target_model`. Note that this scripts runs the model on a device automatically
|
| 260 |
-
provisioned in the cloud. Once the job is submitted, you can navigate to a
|
| 261 |
-
provided job URL to view a variety of on-device performance metrics.
|
| 262 |
-
```python
|
| 263 |
-
profile_job = hub.submit_profile_job(
|
| 264 |
-
model=target_model,
|
| 265 |
-
device=device,
|
| 266 |
-
)
|
| 267 |
-
|
| 268 |
-
```
|
| 269 |
-
|
| 270 |
-
Step 3: **Verify on-device accuracy**
|
| 271 |
-
|
| 272 |
-
To verify the accuracy of the model on-device, you can run on-device inference
|
| 273 |
-
on sample input data on the same cloud hosted device.
|
| 274 |
-
```python
|
| 275 |
-
input_data = torch_model.sample_inputs()
|
| 276 |
-
inference_job = hub.submit_inference_job(
|
| 277 |
-
model=target_model,
|
| 278 |
-
device=device,
|
| 279 |
-
inputs=input_data,
|
| 280 |
-
)
|
| 281 |
-
on_device_output = inference_job.download_output_data()
|
| 282 |
-
|
| 283 |
-
```
|
| 284 |
-
With the output of the model, you can compute like PSNR, relative errors or
|
| 285 |
-
spot check the output with expected output.
|
| 286 |
-
|
| 287 |
-
**Note**: This on-device profiling and inference requires access to Qualcomm®
|
| 288 |
-
AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
| 289 |
-
|
| 290 |
-
|
| 291 |
-
|
| 292 |
-
## Run demo on a cloud-hosted device
|
| 293 |
-
|
| 294 |
-
You can also run the demo on-device.
|
| 295 |
-
|
| 296 |
-
```bash
|
| 297 |
-
python -m qai_hub_models.models.detectron2_detection.demo --eval-mode on-device
|
| 298 |
-
```
|
| 299 |
-
|
| 300 |
-
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
| 301 |
-
environment, please add the following to your cell (instead of the above).
|
| 302 |
-
```
|
| 303 |
-
%run -m qai_hub_models.models.detectron2_detection.demo -- --eval-mode on-device
|
| 304 |
-
```
|
| 305 |
-
|
| 306 |
-
|
| 307 |
-
## Deploying compiled model to Android
|
| 308 |
-
|
| 309 |
-
|
| 310 |
-
The models can be deployed using multiple runtimes:
|
| 311 |
-
- TensorFlow Lite (`.tflite` export): [This
|
| 312 |
-
tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
|
| 313 |
-
guide to deploy the .tflite model in an Android application.
|
| 314 |
-
|
| 315 |
-
|
| 316 |
-
- QNN (`.so` export ): This [sample
|
| 317 |
-
app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
|
| 318 |
-
provides instructions on how to use the `.so` shared library in an Android application.
|
| 319 |
-
|
| 320 |
-
|
| 321 |
-
## View on Qualcomm® AI Hub
|
| 322 |
-
Get more details on Detectron2-Detection's performance across various devices [here](https://aihub.qualcomm.com/models/detectron2_detection).
|
| 323 |
-
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
|
| 324 |
-
|
| 325 |
|
| 326 |
## License
|
| 327 |
* The license for the original implementation of Detectron2-Detection can be found
|
| 328 |
[here](https://github.com/facebookresearch/detectron2/blob/main/LICENSE).
|
| 329 |
|
| 330 |
-
|
| 331 |
-
|
| 332 |
## References
|
| 333 |
* [Source Model Implementation](https://github.com/facebookresearch/detectron2/)
|
| 334 |
|
| 335 |
-
|
| 336 |
-
|
| 337 |
## Community
|
| 338 |
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
|
| 339 |
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
|
| 340 |
-
|
| 341 |
-
|
|
|
|
| 9 |
|
| 10 |
|
| 11 |
|
| 12 |
+
# Detectron2-Detection: Optimized for Qualcomm Devices
|
|
|
|
|
|
|
| 13 |
|
| 14 |
Detectron2-Detection is a machine learning model that can detect objects (trained on COCO dataset).
|
| 15 |
|
| 16 |
+
This is based on the implementation of Detectron2-Detection found [here](https://github.com/facebookresearch/detectron2/).
|
| 17 |
+
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/detectron2_detection) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
|
| 18 |
+
|
| 19 |
+
Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
|
| 20 |
+
|
| 21 |
+
## Getting Started
|
| 22 |
+
There are two ways to deploy this model on your device:
|
| 23 |
+
|
| 24 |
+
### Option 1: Download Pre-Exported Models
|
| 25 |
+
|
| 26 |
+
Below are pre-exported model assets ready for deployment.
|
| 27 |
+
|
| 28 |
+
| Runtime | Precision | Chipset | SDK Versions | Download |
|
| 29 |
+
|---|---|---|---|---|
|
| 30 |
+
| ONNX | w8a16 | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-onnx-w8a16.zip)
|
| 31 |
+
| ONNX | w8a8 | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-onnx-w8a8.zip)
|
| 32 |
+
| QNN_DLC | float | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-qnn_dlc-float.zip)
|
| 33 |
+
| QNN_DLC | w8a16 | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-qnn_dlc-w8a16.zip)
|
| 34 |
+
| QNN_DLC | w8a8 | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-qnn_dlc-w8a8.zip)
|
| 35 |
+
| TFLITE | float | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-tflite-float.zip)
|
| 36 |
+
| TFLITE | w8a8 | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detectron2_detection/releases/v0.46.1/detectron2_detection-tflite-w8a8.zip)
|
| 37 |
+
|
| 38 |
+
For more device-specific assets and performance metrics, visit **[Detectron2-Detection on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/detectron2_detection)**.
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
### Option 2: Export with Custom Configurations
|
| 42 |
+
|
| 43 |
+
Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/detectron2_detection) Python library to compile and export the model with your own:
|
| 44 |
+
- Custom weights (e.g., fine-tuned checkpoints)
|
| 45 |
+
- Custom input shapes
|
| 46 |
+
- Target device and runtime configurations
|
| 47 |
+
|
| 48 |
+
This option is ideal if you need to customize the model beyond the default configuration provided here.
|
| 49 |
+
|
| 50 |
+
See our repository for [Detectron2-Detection on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/detectron2_detection) for usage instructions.
|
| 51 |
+
|
| 52 |
+
## Model Details
|
| 53 |
+
|
| 54 |
+
**Model Type:** Model_use_case.object_detection
|
| 55 |
+
|
| 56 |
+
**Model Stats:**
|
| 57 |
+
- Model checkpoint: faster_rcnn_R_50_C4_1x
|
| 58 |
+
- Input resolution: 800x800
|
| 59 |
+
- Number of parameters: 33.8M
|
| 60 |
+
- Model size (float): 129 MB
|
| 61 |
+
|
| 62 |
+
## Performance Summary
|
| 63 |
+
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|
| 64 |
+
|---|---|---|---|---|---|---
|
| 65 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Snapdragon® X Elite | 20.075 ms | 33 - 33 MB | NPU
|
| 66 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 13.549 ms | 20 - 410 MB | NPU
|
| 67 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Qualcomm® QCS6490 | 2950.562 ms | 185 - 194 MB | CPU
|
| 68 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Qualcomm® QCS9075 | 23.675 ms | 24 - 27 MB | NPU
|
| 69 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 10.616 ms | 14 - 312 MB | NPU
|
| 70 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 1396.35 ms | 186 - 193 MB | CPU
|
| 71 |
+
| Detectron2ProposalGenerator | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 8.65 ms | 0 - 367 MB | NPU
|
| 72 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Snapdragon® X Elite | 12.514 ms | 20 - 20 MB | NPU
|
| 73 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 8.91 ms | 8 - 327 MB | NPU
|
| 74 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Qualcomm® QCS6490 | 443.76 ms | 141 - 175 MB | CPU
|
| 75 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 12.51 ms | 0 - 26 MB | NPU
|
| 76 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Qualcomm® QCS9075 | 12.226 ms | 14 - 17 MB | NPU
|
| 77 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 7.07 ms | 8 - 261 MB | NPU
|
| 78 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 353.326 ms | 178 - 186 MB | CPU
|
| 79 |
+
| Detectron2ProposalGenerator | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.679 ms | 0 - 253 MB | NPU
|
| 80 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Snapdragon® X Elite | 57.369 ms | 7 - 7 MB | NPU
|
| 81 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 39.993 ms | 0 - 551 MB | NPU
|
| 82 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 188.472 ms | 2 - 460 MB | NPU
|
| 83 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 57.056 ms | 7 - 10 MB | NPU
|
| 84 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® SA8775P | 70.835 ms | 1 - 459 MB | NPU
|
| 85 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® QCS9075 | 84.478 ms | 7 - 27 MB | NPU
|
| 86 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 90.368 ms | 7 - 500 MB | NPU
|
| 87 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® SA7255P | 188.472 ms | 2 - 460 MB | NPU
|
| 88 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Qualcomm® SA8295P | 80.749 ms | 0 - 422 MB | NPU
|
| 89 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 29.479 ms | 7 - 473 MB | NPU
|
| 90 |
+
| Detectron2ProposalGenerator | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 23.267 ms | 0 - 470 MB | NPU
|
| 91 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Snapdragon® X Elite | 69.762 ms | 4 - 4 MB | NPU
|
| 92 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 45.848 ms | 0 - 547 MB | NPU
|
| 93 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 162.726 ms | 3 - 14 MB | NPU
|
| 94 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 113.95 ms | 1 - 449 MB | NPU
|
| 95 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 69.053 ms | 4 - 241 MB | NPU
|
| 96 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® SA8775P | 69.057 ms | 2 - 450 MB | NPU
|
| 97 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 77.265 ms | 0 - 10 MB | NPU
|
| 98 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 448.052 ms | 4 - 561 MB | NPU
|
| 99 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 80.286 ms | 4 - 525 MB | NPU
|
| 100 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® SA7255P | 113.95 ms | 1 - 449 MB | NPU
|
| 101 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Qualcomm® SA8295P | 78.49 ms | 0 - 445 MB | NPU
|
| 102 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 25.412 ms | 4 - 428 MB | NPU
|
| 103 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 55.288 ms | 4 - 504 MB | NPU
|
| 104 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 22.808 ms | 4 - 478 MB | NPU
|
| 105 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Snapdragon® X Elite | 40.367 ms | 2 - 2 MB | NPU
|
| 106 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.673 ms | 2 - 493 MB | NPU
|
| 107 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 65.487 ms | 1 - 7 MB | NPU
|
| 108 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 69.69 ms | 2 - 427 MB | NPU
|
| 109 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 39.174 ms | 2 - 6 MB | NPU
|
| 110 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® SA8775P | 39.385 ms | 2 - 428 MB | NPU
|
| 111 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 44.448 ms | 0 - 6 MB | NPU
|
| 112 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 181.15 ms | 2 - 416 MB | NPU
|
| 113 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 46.802 ms | 2 - 483 MB | NPU
|
| 114 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® SA7255P | 69.69 ms | 2 - 427 MB | NPU
|
| 115 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Qualcomm® SA8295P | 45.548 ms | 2 - 428 MB | NPU
|
| 116 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 22.957 ms | 2 - 414 MB | NPU
|
| 117 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 30.021 ms | 2 - 434 MB | NPU
|
| 118 |
+
| Detectron2ProposalGenerator | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 21.186 ms | 2 - 420 MB | NPU
|
| 119 |
+
| Detectron2ProposalGenerator | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 30.494 ms | 9 - 564 MB | NPU
|
| 120 |
+
| Detectron2ProposalGenerator | TFLITE | float | Qualcomm® SA8295P | 62.075 ms | 11 - 443 MB | NPU
|
| 121 |
+
| Detectron2ProposalGenerator | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 23.447 ms | 9 - 474 MB | NPU
|
| 122 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 12.583 ms | 2 - 469 MB | NPU
|
| 123 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 30.168 ms | 3 - 400 MB | NPU
|
| 124 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 17.066 ms | 3 - 6 MB | NPU
|
| 125 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® SA8775P | 17.019 ms | 3 - 399 MB | NPU
|
| 126 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® QCS9075 | 21.334 ms | 1 - 29 MB | NPU
|
| 127 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 21.877 ms | 3 - 455 MB | NPU
|
| 128 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® SA7255P | 30.168 ms | 3 - 400 MB | NPU
|
| 129 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Qualcomm® SA8295P | 21.667 ms | 3 - 400 MB | NPU
|
| 130 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 8.956 ms | 1 - 421 MB | NPU
|
| 131 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 23.739 ms | 2 - 441 MB | NPU
|
| 132 |
+
| Detectron2ProposalGenerator | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 7.418 ms | 1 - 420 MB | NPU
|
| 133 |
+
| Detectron2ROIHead | ONNX | w8a16 | Snapdragon® X Elite | 255.728 ms | 1059 - 1059 MB | NPU
|
| 134 |
+
| Detectron2ROIHead | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 261.735 ms | 955 - 1381 MB | NPU
|
| 135 |
+
| Detectron2ROIHead | ONNX | w8a16 | Qualcomm® QCS6490 | 8775.362 ms | 3037 - 3052 MB | CPU
|
| 136 |
+
| Detectron2ROIHead | ONNX | w8a16 | Qualcomm® QCS9075 | 325.2 ms | 1014 - 1017 MB | NPU
|
| 137 |
+
| Detectron2ROIHead | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 241.107 ms | 986 - 1368 MB | NPU
|
| 138 |
+
| Detectron2ROIHead | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 3600.084 ms | 2946 - 2953 MB | CPU
|
| 139 |
+
| Detectron2ROIHead | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 188.327 ms | 994 - 1380 MB | NPU
|
| 140 |
+
| Detectron2ROIHead | ONNX | w8a8 | Snapdragon® X Elite | 98.706 ms | 530 - 530 MB | NPU
|
| 141 |
+
| Detectron2ROIHead | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 91.998 ms | 507 - 844 MB | NPU
|
| 142 |
+
| Detectron2ROIHead | ONNX | w8a8 | Qualcomm® QCS6490 | 1718.929 ms | 2531 - 2548 MB | CPU
|
| 143 |
+
| Detectron2ROIHead | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 113.12 ms | 507 - 512 MB | NPU
|
| 144 |
+
| Detectron2ROIHead | ONNX | w8a8 | Qualcomm® QCS9075 | 131.146 ms | 507 - 510 MB | NPU
|
| 145 |
+
| Detectron2ROIHead | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 87.37 ms | 505 - 800 MB | NPU
|
| 146 |
+
| Detectron2ROIHead | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 988.134 ms | 2432 - 2438 MB | CPU
|
| 147 |
+
| Detectron2ROIHead | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 68.296 ms | 408 - 704 MB | NPU
|
| 148 |
+
| Detectron2ROIHead | QNN_DLC | float | Snapdragon® X Elite | 64.387 ms | 10 - 10 MB | NPU
|
| 149 |
+
| Detectron2ROIHead | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 49.149 ms | 10 - 2078 MB | NPU
|
| 150 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 332.997 ms | 1 - 1470 MB | NPU
|
| 151 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 65.223 ms | 10 - 14 MB | NPU
|
| 152 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® SA8775P | 96.56 ms | 1 - 1470 MB | NPU
|
| 153 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® QCS9075 | 115.613 ms | 10 - 21 MB | NPU
|
| 154 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 119.198 ms | 10 - 449 MB | NPU
|
| 155 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® SA7255P | 332.997 ms | 1 - 1470 MB | NPU
|
| 156 |
+
| Detectron2ROIHead | QNN_DLC | float | Qualcomm® SA8295P | 88.151 ms | 0 - 391 MB | NPU
|
| 157 |
+
| Detectron2ROIHead | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 42.17 ms | 10 - 1472 MB | NPU
|
| 158 |
+
| Detectron2ROIHead | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 27.391 ms | 10 - 1559 MB | NPU
|
| 159 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Snapdragon® X Elite | 33.565 ms | 5 - 5 MB | NPU
|
| 160 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 24.348 ms | 9 - 439 MB | NPU
|
| 161 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 147.668 ms | 4 - 10 MB | NPU
|
| 162 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 98.919 ms | 5 - 371 MB | NPU
|
| 163 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 32.776 ms | 5 - 7 MB | NPU
|
| 164 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® SA8775P | 32.521 ms | 4 - 371 MB | NPU
|
| 165 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 44.613 ms | 4 - 11 MB | NPU
|
| 166 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 673.967 ms | 5 - 558 MB | NPU
|
| 167 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 61.039 ms | 5 - 434 MB | NPU
|
| 168 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® SA7255P | 98.919 ms | 5 - 371 MB | NPU
|
| 169 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Qualcomm® SA8295P | 51.421 ms | 0 - 368 MB | NPU
|
| 170 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 19.711 ms | 5 - 423 MB | NPU
|
| 171 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 56.396 ms | 5 - 463 MB | NPU
|
| 172 |
+
| Detectron2ROIHead | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 13.49 ms | 5 - 426 MB | NPU
|
| 173 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Snapdragon® X Elite | 19.655 ms | 2 - 2 MB | NPU
|
| 174 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 14.052 ms | 2 - 360 MB | NPU
|
| 175 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 85.551 ms | 2 - 6 MB | NPU
|
| 176 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 53.032 ms | 0 - 305 MB | NPU
|
| 177 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 19.036 ms | 2 - 5 MB | NPU
|
| 178 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® SA8775P | 88.62 ms | 2 - 306 MB | NPU
|
| 179 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 80.739 ms | 0 - 4 MB | NPU
|
| 180 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 322.671 ms | 2 - 397 MB | NPU
|
| 181 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 40.291 ms | 2 - 353 MB | NPU
|
| 182 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® SA7255P | 53.032 ms | 0 - 305 MB | NPU
|
| 183 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Qualcomm® SA8295P | 29.193 ms | 2 - 305 MB | NPU
|
| 184 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 10.347 ms | 2 - 319 MB | NPU
|
| 185 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 31.192 ms | 2 - 337 MB | NPU
|
| 186 |
+
| Detectron2ROIHead | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 8.385 ms | 2 - 320 MB | NPU
|
| 187 |
+
| Detectron2ROIHead | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 846.324 ms | 0 - 3351 MB | CPU
|
| 188 |
+
| Detectron2ROIHead | TFLITE | float | Qualcomm® SA8295P | 1221.27 ms | 0 - 1587 MB | CPU
|
| 189 |
+
| Detectron2ROIHead | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 648.238 ms | 1 - 2768 MB | CPU
|
| 190 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 250.155 ms | 0 - 2305 MB | NPU
|
| 191 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 502.622 ms | 0 - 1304 MB | NPU
|
| 192 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 310.101 ms | 0 - 3 MB | NPU
|
| 193 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® SA8775P | 347.063 ms | 0 - 1305 MB | NPU
|
| 194 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® QCS9075 | 359.401 ms | 0 - 33 MB | NPU
|
| 195 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 799.77 ms | 0 - 845 MB | NPU
|
| 196 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® SA7255P | 502.622 ms | 0 - 1304 MB | NPU
|
| 197 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Qualcomm® SA8295P | 374.253 ms | 0 - 735 MB | NPU
|
| 198 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 172.126 ms | 0 - 1305 MB | NPU
|
| 199 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1822.304 ms | 1118 - 2387 MB | NPU
|
| 200 |
+
| Detectron2ROIHead | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 162.543 ms | 0 - 1612 MB | NPU
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 201 |
|
| 202 |
## License
|
| 203 |
* The license for the original implementation of Detectron2-Detection can be found
|
| 204 |
[here](https://github.com/facebookresearch/detectron2/blob/main/LICENSE).
|
| 205 |
|
|
|
|
|
|
|
| 206 |
## References
|
| 207 |
* [Source Model Implementation](https://github.com/facebookresearch/detectron2/)
|
| 208 |
|
|
|
|
|
|
|
| 209 |
## Community
|
| 210 |
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
|
| 211 |
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
|
|
|
|
|
|
tool-versions.yaml
DELETED
|
@@ -1,4 +0,0 @@
|
|
| 1 |
-
tool_versions:
|
| 2 |
-
onnx:
|
| 3 |
-
qairt: 2.37.1.250807093845_124904
|
| 4 |
-
onnx_runtime: 1.23.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|