Hugging Face's logo

qualcomm
/
Yolo-v7

Object Detection
PyTorch
bu_auto
real_time
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
231077d
·
verified ·
1 Parent(s): f36f8e6

v0.46.1

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.46.1 for changelog.

Files changed (1) hide show
  1. README.md +117 -286
README.md CHANGED
@@ -11,304 +11,135 @@ pipeline_tag: object-detection
11
 
12
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolov7/web-assets/model_demo.png)
13
 
14
- # Yolo-v7: Optimized for Mobile Deployment
15
- ## Real-time object detection optimized for mobile and edge
16
-
17
 
18
  YoloV7 is a machine learning model that predicts bounding boxes and classes of objects in an image.
19
 
20
- This model is an implementation of Yolo-v7 found [here](https://github.com/WongKinYiu/yolov7/).
21
-
22
-
23
- This repository provides scripts to run Yolo-v7 on Qualcomm® devices.
24
- More details on model performance across various devices, can be found
25
- [here](https://aihub.qualcomm.com/models/yolov7).
26
-
27
- **WARNING**: The model assets are not readily available for download due to licensing restrictions.
28
-
29
- ### Model Details
30
-
31
- - **Model Type:** Model_use_case.object_detection
32
- - **Model Stats:**
33
- - Model checkpoint: YoloV7 Tiny
34
- - Input resolution: 640x640
35
- - Number of parameters: 6.24M
36
- - Model size (float): 23.8 MB
37
- - Model size (w8a8): 6.23 MB
38
- - Model size (w8a16): 6.66 MB
39
-
40
- | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
41
- |---|---|---|---|---|---|---|---|---|
42
- | Yolo-v7 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 29.023 ms | 1 - 135 MB | NPU | -- |
43
- | Yolo-v7 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 18.738 ms | 0 - 267 MB | NPU | -- |
44
- | Yolo-v7 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 22.496 ms | 1 - 173 MB | NPU | -- |
45
- | Yolo-v7 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 10.64 ms | 5 - 165 MB | NPU | -- |
46
- | Yolo-v7 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 13.977 ms | 1 - 3 MB | NPU | -- |
47
- | Yolo-v7 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 6.464 ms | 5 - 7 MB | NPU | -- |
48
- | Yolo-v7 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 8.321 ms | 0 - 12 MB | NPU | -- |
49
- | Yolo-v7 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 15.564 ms | 1 - 137 MB | NPU | -- |
50
- | Yolo-v7 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 8.118 ms | 1 - 257 MB | NPU | -- |
51
- | Yolo-v7 | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 29.023 ms | 1 - 135 MB | NPU | -- |
52
- | Yolo-v7 | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 18.738 ms | 0 - 267 MB | NPU | -- |
53
- | Yolo-v7 | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 18.452 ms | 1 - 145 MB | NPU | -- |
54
- | Yolo-v7 | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 9.217 ms | 0 - 147 MB | NPU | -- |
55
- | Yolo-v7 | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 15.564 ms | 1 - 137 MB | NPU | -- |
56
- | Yolo-v7 | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 8.118 ms | 1 - 257 MB | NPU | -- |
57
- | Yolo-v7 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 10.814 ms | 0 - 165 MB | NPU | -- |
58
- | Yolo-v7 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 4.549 ms | 5 - 491 MB | NPU | -- |
59
- | Yolo-v7 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 5.621 ms | 6 - 461 MB | NPU | -- |
60
- | Yolo-v7 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 16.649 ms | 0 - 135 MB | GPU | -- |
61
- | Yolo-v7 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 3.539 ms | 0 - 237 MB | NPU | -- |
62
- | Yolo-v7 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 5.363 ms | 3 - 224 MB | NPU | -- |
63
- | Yolo-v7 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 6.522 ms | 1 - 145 MB | NPU | -- |
64
- | Yolo-v7 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 2.875 ms | 5 - 263 MB | NPU | -- |
65
- | Yolo-v7 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 3.64 ms | 0 - 224 MB | NPU | -- |
66
- | Yolo-v7 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 6.891 ms | 5 - 5 MB | NPU | -- |
67
- | Yolo-v7 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 8.47 ms | 9 - 9 MB | NPU | -- |
68
- | Yolo-v7 | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | QNN_DLC | 31.878 ms | 2 - 250 MB | NPU | -- |
69
- | Yolo-v7 | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | ONNX | 215.545 ms | 83 - 97 MB | CPU | -- |
70
- | Yolo-v7 | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 11.435 ms | 2 - 6 MB | NPU | -- |
71
- | Yolo-v7 | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 437.653 ms | 83 - 88 MB | CPU | -- |
72
- | Yolo-v7 | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 8.991 ms | 2 - 299 MB | NPU | -- |
73
- | Yolo-v7 | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 6.094 ms | 2 - 331 MB | NPU | -- |
74
- | Yolo-v7 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.927 ms | 2 - 5 MB | NPU | -- |
75
- | Yolo-v7 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 5.196 ms | 0 - 7 MB | NPU | -- |
76
- | Yolo-v7 | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 4.563 ms | 0 - 301 MB | NPU | -- |
77
- | Yolo-v7 | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 8.991 ms | 2 - 299 MB | NPU | -- |
78
- | Yolo-v7 | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 5.677 ms | 0 - 299 MB | NPU | -- |
79
- | Yolo-v7 | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 4.563 ms | 0 - 301 MB | NPU | -- |
80
- | Yolo-v7 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.818 ms | 2 - 328 MB | NPU | -- |
81
- | Yolo-v7 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 3.481 ms | 3 - 327 MB | NPU | -- |
82
- | Yolo-v7 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 2.061 ms | 2 - 304 MB | NPU | -- |
83
- | Yolo-v7 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 2.67 ms | 0 - 283 MB | NPU | -- |
84
- | Yolo-v7 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 5.027 ms | 2 - 233 MB | NPU | -- |
85
- | Yolo-v7 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 201.854 ms | 82 - 99 MB | CPU | -- |
86
- | Yolo-v7 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 1.613 ms | 2 - 298 MB | NPU | -- |
87
- | Yolo-v7 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 2.143 ms | 0 - 283 MB | NPU | -- |
88
- | Yolo-v7 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 4.432 ms | 2 - 2 MB | NPU | -- |
89
- | Yolo-v7 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.419 ms | 5 - 5 MB | NPU | -- |
90
- | Yolo-v7 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | TFLITE | 14.407 ms | 0 - 148 MB | NPU | -- |
91
- | Yolo-v7 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | QNN_DLC | 14.621 ms | 1 - 167 MB | NPU | -- |
92
- | Yolo-v7 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | ONNX | 54.455 ms | 39 - 55 MB | CPU | -- |
93
- | Yolo-v7 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | TFLITE | 6.113 ms | 0 - 9 MB | NPU | -- |
94
- | Yolo-v7 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 6.251 ms | 3 - 6 MB | NPU | -- |
95
- | Yolo-v7 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 74.368 ms | 38 - 45 MB | CPU | -- |
96
- | Yolo-v7 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 4.74 ms | 0 - 154 MB | NPU | -- |
97
- | Yolo-v7 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 4.592 ms | 1 - 147 MB | NPU | -- |
98
- | Yolo-v7 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 3.118 ms | 0 - 162 MB | NPU | -- |
99
- | Yolo-v7 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 2.863 ms | 1 - 174 MB | NPU | -- |
100
- | Yolo-v7 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 2.237 ms | 0 - 2 MB | NPU | -- |
101
- | Yolo-v7 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 2.139 ms | 1 - 4 MB | NPU | -- |
102
- | Yolo-v7 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 2.708 ms | 0 - 8 MB | NPU | -- |
103
- | Yolo-v7 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 2.67 ms | 0 - 156 MB | NPU | -- |
104
- | Yolo-v7 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 2.523 ms | 1 - 148 MB | NPU | -- |
105
- | Yolo-v7 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 4.74 ms | 0 - 154 MB | NPU | -- |
106
- | Yolo-v7 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 4.592 ms | 1 - 147 MB | NPU | -- |
107
- | Yolo-v7 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 3.703 ms | 0 - 146 MB | NPU | -- |
108
- | Yolo-v7 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 3.5 ms | 0 - 155 MB | NPU | -- |
109
- | Yolo-v7 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 2.67 ms | 0 - 156 MB | NPU | -- |
110
- | Yolo-v7 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 2.523 ms | 1 - 148 MB | NPU | -- |
111
- | Yolo-v7 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1.424 ms | 0 - 178 MB | NPU | -- |
112
- | Yolo-v7 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.417 ms | 1 - 171 MB | NPU | -- |
113
- | Yolo-v7 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.744 ms | 0 - 159 MB | NPU | -- |
114
- | Yolo-v7 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 1.214 ms | 0 - 161 MB | NPU | -- |
115
- | Yolo-v7 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 1.053 ms | 1 - 152 MB | NPU | -- |
116
- | Yolo-v7 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 1.391 ms | 0 - 135 MB | NPU | -- |
117
- | Yolo-v7 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 2.636 ms | 0 - 148 MB | NPU | -- |
118
- | Yolo-v7 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 2.643 ms | 1 - 296 MB | NPU | -- |
119
- | Yolo-v7 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 50.926 ms | 41 - 58 MB | CPU | -- |
120
- | Yolo-v7 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 0.982 ms | 0 - 176 MB | NPU | -- |
121
- | Yolo-v7 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 0.895 ms | 1 - 151 MB | NPU | -- |
122
- | Yolo-v7 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 1.245 ms | 0 - 135 MB | NPU | -- |
123
- | Yolo-v7 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 2.397 ms | 1 - 1 MB | NPU | -- |
124
- | Yolo-v7 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.752 ms | 5 - 5 MB | NPU | -- |
125
-
126
-
127
-
128
-
129
- ## Installation
130
-
131
-
132
- Install the package via pip:
133
- ```bash
134
- # NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
135
- pip install "qai-hub-models[yolov7]"
136
- ```
137
-
138
-
139
- ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
140
-
141
- Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
142
- Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
143
-
144
- With this API token, you can configure your client to run models on the cloud
145
- hosted devices.
146
- ```bash
147
- qai-hub configure --api_token API_TOKEN
148
- ```
149
- Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
150
-
151
-
152
-
153
- ## Demo off target
154
-
155
- The package contains a simple end-to-end demo that downloads pre-trained
156
- weights and runs this model on a sample input.
157
-
158
- ```bash
159
- python -m qai_hub_models.models.yolov7.demo
160
- ```
161
-
162
- The above demo runs a reference implementation of pre-processing, model
163
- inference, and post processing.
164
-
165
- **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
166
- environment, please add the following to your cell (instead of the above).
167
- ```
168
- %run -m qai_hub_models.models.yolov7.demo
169
- ```
170
-
171
-
172
- ### Run model on a cloud-hosted device
173
-
174
- In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
175
- device. This script does the following:
176
- * Performance check on-device on a cloud-hosted device
177
- * Downloads compiled assets that can be deployed on-device for Android.
178
- * Accuracy check between PyTorch and on-device outputs.
179
-
180
- ```bash
181
- python -m qai_hub_models.models.yolov7.export
182
- ```
183
-
184
-
185
-
186
- ## How does this work?
187
-
188
- This [export script](https://aihub.qualcomm.com/models/yolov7/qai_hub_models/models/Yolo-v7/export.py)
189
- leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
190
- on-device. Lets go through each step below in detail:
191
-
192
- Step 1: **Compile model for on-device deployment**
193
-
194
- To compile a PyTorch model for on-device deployment, we first trace the model
195
- in memory using the `jit.trace` and then call the `submit_compile_job` API.
196
-
197
- ```python
198
- import torch
199
-
200
- import qai_hub as hub
201
- from qai_hub_models.models.yolov7 import Model
202
-
203
- # Load the model
204
- torch_model = Model.from_pretrained()
205
-
206
- # Device
207
- device = hub.Device("Samsung Galaxy S25")
208
-
209
- # Trace model
210
- input_shape = torch_model.get_input_spec()
211
- sample_inputs = torch_model.sample_inputs()
212
-
213
- pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
214
-
215
- # Compile model on a specific device
216
- compile_job = hub.submit_compile_job(
217
- model=pt_model,
218
- device=device,
219
- input_specs=torch_model.get_input_spec(),
220
- )
221
-
222
- # Get target model to run on-device
223
- target_model = compile_job.get_target_model()
224
-
225
- ```
226
-
227
-
228
- Step 2: **Performance profiling on cloud-hosted device**
229
-
230
- After compiling models from step 1. Models can be profiled model on-device using the
231
- `target_model`. Note that this scripts runs the model on a device automatically
232
- provisioned in the cloud. Once the job is submitted, you can navigate to a
233
- provided job URL to view a variety of on-device performance metrics.
234
- ```python
235
- profile_job = hub.submit_profile_job(
236
- model=target_model,
237
- device=device,
238
- )
239
-
240
- ```
241
-
242
- Step 3: **Verify on-device accuracy**
243
-
244
- To verify the accuracy of the model on-device, you can run on-device inference
245
- on sample input data on the same cloud hosted device.
246
- ```python
247
- input_data = torch_model.sample_inputs()
248
- inference_job = hub.submit_inference_job(
249
- model=target_model,
250
- device=device,
251
- inputs=input_data,
252
- )
253
- on_device_output = inference_job.download_output_data()
254
-
255
- ```
256
- With the output of the model, you can compute like PSNR, relative errors or
257
- spot check the output with expected output.
258
-
259
- **Note**: This on-device profiling and inference requires access to Qualcomm®
260
- AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
261
-
262
-
263
-
264
- ## Run demo on a cloud-hosted device
265
-
266
- You can also run the demo on-device.
267
-
268
- ```bash
269
- python -m qai_hub_models.models.yolov7.demo --eval-mode on-device
270
- ```
271
-
272
- **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
273
- environment, please add the following to your cell (instead of the above).
274
- ```
275
- %run -m qai_hub_models.models.yolov7.demo -- --eval-mode on-device
276
- ```
277
-
278
-
279
- ## Deploying compiled model to Android
280
-
281
-
282
- The models can be deployed using multiple runtimes:
283
- - TensorFlow Lite (`.tflite` export): [This
284
- tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
285
- guide to deploy the .tflite model in an Android application.
286
-
287
-
288
- - QNN (`.so` export ): This [sample
289
- app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
290
- provides instructions on how to use the `.so` shared library in an Android application.
291
-
292
-
293
- ## View on Qualcomm® AI Hub
294
- Get more details on Yolo-v7's performance across various devices [here](https://aihub.qualcomm.com/models/yolov7).
295
- Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
296
-
297
 
298
  ## License
299
  * The license for the original implementation of Yolo-v7 can be found
300
  [here](https://github.com/WongKinYiu/yolov7/blob/main/LICENSE.md).
301
 
302
-
303
-
304
  ## References
305
  * [YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors](https://arxiv.org/abs/2207.02696)
306
  * [Source Model Implementation](https://github.com/WongKinYiu/yolov7/)
307
 
308
-
309
-
310
  ## Community
311
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
312
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
313
-
314
-
 
11
 
12
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolov7/web-assets/model_demo.png)
13
 
14
+ # Yolo-v7: Optimized for Qualcomm Devices
 
 
15
 
16
  YoloV7 is a machine learning model that predicts bounding boxes and classes of objects in an image.
17
 
18
+ This is based on the implementation of Yolo-v7 found [here](https://github.com/WongKinYiu/yolov7/).
19
+ 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/yolov7) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
20
+
21
+ 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.
22
+
23
+ ## Getting Started
24
+ Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
25
+ Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov7) Python library to compile and export the model with your own:
26
+ - Custom weights (e.g., fine-tuned checkpoints)
27
+ - Custom input shapes
28
+ - Target device and runtime configurations
29
+
30
+ See our repository for [Yolo-v7 on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov7) for usage instructions.
31
+
32
+
33
+ ## Model Details
34
+
35
+ **Model Type:** Model_use_case.object_detection
36
+
37
+ **Model Stats:**
38
+ - Model checkpoint: YoloV7 Tiny
39
+ - Input resolution: 640x640
40
+ - Number of parameters: 6.24M
41
+ - Model size (float): 23.8 MB
42
+ - Model size (w8a8): 6.23 MB
43
+ - Model size (w8a16): 6.66 MB
44
+
45
+ ## Performance Summary
46
+ | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
47
+ |---|---|---|---|---|---|---
48
+ | Yolo-v7 | ONNX | float | Snapdragon® X Elite | 11.088 ms | 8 - 8 MB | NPU
49
+ | Yolo-v7 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 8.128 ms | 6 - 142 MB | NPU
50
+ | Yolo-v7 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 11.403 ms | 0 - 12 MB | NPU
51
+ | Yolo-v7 | ONNX | float | Qualcomm® QCS9075 | 11.665 ms | 5 - 7 MB | NPU
52
+ | Yolo-v7 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.697 ms | 1 - 113 MB | NPU
53
+ | Yolo-v7 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.376 ms | 5 - 127 MB | NPU
54
+ | Yolo-v7 | ONNX | w8a16 | Snapdragon® X Elite | 5.449 ms | 4 - 4 MB | NPU
55
+ | Yolo-v7 | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 3.584 ms | 0 - 307 MB | NPU
56
+ | Yolo-v7 | ONNX | w8a16 | Qualcomm® QCS6490 | 425.947 ms | 85 - 90 MB | CPU
57
+ | Yolo-v7 | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 5.19 ms | 0 - 11 MB | NPU
58
+ | Yolo-v7 | ONNX | w8a16 | Qualcomm® QCS9075 | 6.652 ms | 2 - 5 MB | NPU
59
+ | Yolo-v7 | ONNX | w8a16 | Qualcomm® QCM6690 | 215.3 ms | 83 - 92 MB | CPU
60
+ | Yolo-v7 | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.683 ms | 0 - 281 MB | NPU
61
+ | Yolo-v7 | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 188.489 ms | 83 - 91 MB | CPU
62
+ | Yolo-v7 | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.422 ms | 0 - 284 MB | NPU
63
+ | Yolo-v7 | ONNX | w8a8 | Snapdragon® X Elite | 2.743 ms | 5 - 5 MB | NPU
64
+ | Yolo-v7 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.779 ms | 0 - 158 MB | NPU
65
+ | Yolo-v7 | ONNX | w8a8 | Qualcomm® QCS6490 | 80.342 ms | 39 - 47 MB | CPU
66
+ | Yolo-v7 | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 2.767 ms | 0 - 46 MB | NPU
67
+ | Yolo-v7 | ONNX | w8a8 | Qualcomm® QCS9075 | 3.581 ms | 1 - 4 MB | NPU
68
+ | Yolo-v7 | ONNX | w8a8 | Qualcomm® QCM6690 | 54.002 ms | 39 - 49 MB | CPU
69
+ | Yolo-v7 | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.414 ms | 0 - 138 MB | NPU
70
+ | Yolo-v7 | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 42.962 ms | 40 - 49 MB | CPU
71
+ | Yolo-v7 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.246 ms | 0 - 132 MB | NPU
72
+ | Yolo-v7 | QNN_DLC | float | Snapdragon® X Elite | 7.55 ms | 5 - 5 MB | NPU
73
+ | Yolo-v7 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 5.474 ms | 0 - 179 MB | NPU
74
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 19.823 ms | 0 - 153 MB | NPU
75
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 7.319 ms | 5 - 96 MB | NPU
76
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® SA8775P | 8.931 ms | 1 - 158 MB | NPU
77
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® QCS9075 | 9.147 ms | 7 - 13 MB | NPU
78
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 11.338 ms | 5 - 183 MB | NPU
79
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® SA7255P | 19.823 ms | 0 - 153 MB | NPU
80
+ | Yolo-v7 | QNN_DLC | float | Qualcomm® SA8295P | 10.454 ms | 0 - 151 MB | NPU
81
+ | Yolo-v7 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.337 ms | 0 - 158 MB | NPU
82
+ | Yolo-v7 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.807 ms | 5 - 162 MB | NPU
83
+ | Yolo-v7 | QNN_DLC | w8a16 | Snapdragon® X Elite | 4.5 ms | 2 - 2 MB | NPU
84
+ | Yolo-v7 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.887 ms | 2 - 300 MB | NPU
85
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 11.637 ms | 1 - 6 MB | NPU
86
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 9.503 ms | 1 - 270 MB | NPU
87
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 4.11 ms | 2 - 5 MB | NPU
88
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® SA8775P | 4.72 ms | 0 - 272 MB | NPU
89
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 4.809 ms | 2 - 6 MB | NPU
90
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 32.711 ms | 2 - 228 MB | NPU
91
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 6.912 ms | 2 - 307 MB | NPU
92
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® SA7255P | 9.503 ms | 1 - 270 MB | NPU
93
+ | Yolo-v7 | QNN_DLC | w8a16 | Qualcomm® SA8295P | 5.93 ms | 0 - 274 MB | NPU
94
+ | Yolo-v7 | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.182 ms | 2 - 274 MB | NPU
95
+ | Yolo-v7 | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 5.026 ms | 2 - 217 MB | NPU
96
+ | Yolo-v7 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.694 ms | 2 - 275 MB | NPU
97
+ | Yolo-v7 | QNN_DLC | w8a8 | Snapdragon® X Elite | 2.449 ms | 1 - 1 MB | NPU
98
+ | Yolo-v7 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.4 ms | 1 - 72 MB | NPU
99
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 6.165 ms | 3 - 6 MB | NPU
100
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 4.724 ms | 1 - 53 MB | NPU
101
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 2.169 ms | 1 - 3 MB | NPU
102
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 2.601 ms | 1 - 55 MB | NPU
103
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 2.478 ms | 1 - 4 MB | NPU
104
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 13.831 ms | 1 - 174 MB | NPU
105
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 2.827 ms | 1 - 72 MB | NPU
106
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® SA7255P | 4.724 ms | 1 - 53 MB | NPU
107
+ | Yolo-v7 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 3.6 ms | 1 - 51 MB | NPU
108
+ | Yolo-v7 | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.08 ms | 1 - 55 MB | NPU
109
+ | Yolo-v7 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2.686 ms | 1 - 263 MB | NPU
110
+ | Yolo-v7 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.863 ms | 1 - 54 MB | NPU
111
+ | Yolo-v7 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.455 ms | 1 - 185 MB | NPU
112
+ | Yolo-v7 | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 20.204 ms | 1 - 157 MB | NPU
113
+ | Yolo-v7 | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 7.551 ms | 1 - 2 MB | NPU
114
+ | Yolo-v7 | TFLITE | float | Qualcomm® SA8775P | 9.167 ms | 1 - 160 MB | NPU
115
+ | Yolo-v7 | TFLITE | float | Qualcomm® QCS9075 | 9.243 ms | 1 - 20 MB | NPU
116
+ | Yolo-v7 | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 11.373 ms | 1 - 193 MB | NPU
117
+ | Yolo-v7 | TFLITE | float | Qualcomm® SA7255P | 20.204 ms | 1 - 157 MB | NPU
118
+ | Yolo-v7 | TFLITE | float | Qualcomm® SA8295P | 10.411 ms | 1 - 158 MB | NPU
119
+ | Yolo-v7 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.35 ms | 1 - 166 MB | NPU
120
+ | Yolo-v7 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.807 ms | 0 - 163 MB | NPU
121
+ | Yolo-v7 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.387 ms | 0 - 86 MB | NPU
122
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCS6490 | 6.147 ms | 0 - 10 MB | NPU
123
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 4.689 ms | 0 - 66 MB | NPU
124
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 2.154 ms | 0 - 3 MB | NPU
125
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® SA8775P | 10.921 ms | 0 - 65 MB | NPU
126
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCS9075 | 2.443 ms | 0 - 9 MB | NPU
127
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCM6690 | 14.75 ms | 0 - 161 MB | NPU
128
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 3.006 ms | 0 - 63 MB | NPU
129
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® SA7255P | 4.689 ms | 0 - 66 MB | NPU
130
+ | Yolo-v7 | TFLITE | w8a8 | Qualcomm® SA8295P | 3.726 ms | 0 - 43 MB | NPU
131
+ | Yolo-v7 | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.212 ms | 0 - 66 MB | NPU
132
+ | Yolo-v7 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2.637 ms | 0 - 171 MB | NPU
133
+ | Yolo-v7 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.988 ms | 0 - 84 MB | NPU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
134
 
135
  ## License
136
  * The license for the original implementation of Yolo-v7 can be found
137
  [here](https://github.com/WongKinYiu/yolov7/blob/main/LICENSE.md).
138
 
 
 
139
  ## References
140
  * [YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors](https://arxiv.org/abs/2207.02696)
141
  * [Source Model Implementation](https://github.com/WongKinYiu/yolov7/)
142
 
 
 
143
  ## Community
144
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
145
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).