Hugging Face's logo

qualcomm
/
EfficientNet-B4

Image Classification
PyTorch
backbone
bu_auto
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
cd70384
·
verified ·
1 Parent(s): 141111a

v0.46.1

Browse files

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

Files changed (6) hide show
  1. EfficientNet-B4_float.dlc +0 -3
  2. EfficientNet-B4_float.onnx.zip +0 -3
  3. EfficientNet-B4_float.tflite +0 -3
  4. EfficientNet-B4_w8a16.dlc +0 -3
  5. README.md +81 -232
  6. tool-versions.yaml +0 -3
EfficientNet-B4_float.dlc DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:39ae115eae8ad7c7c8f058aad32ce3e6ea25e39e247059521e6c5a837f834034
3
- size 77623708
 
 
 
 
EfficientNet-B4_float.onnx.zip DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:c990f2c599b6a0bea9b1f9bea028865e58af998037fd1115797a653b7c99a782
3
- size 72008241
 
 
 
 
EfficientNet-B4_float.tflite DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:c2880b4c138b09c2bc61b73c968d697a07cd2fab0dc84b7058eb0b1c138f3b0f
3
- size 77222888
 
 
 
 
EfficientNet-B4_w8a16.dlc DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:76531ff401823a95c4f8e970f33ed97d4d16cc80f1c23537526c7cec25bef580
3
- size 25228468
 
 
 
 
README.md CHANGED
@@ -11,250 +11,99 @@ pipeline_tag: image-classification
11
 
12
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/web-assets/model_demo.png)
13
 
14
- # EfficientNet-B4: Optimized for Mobile Deployment
15
- ## Imagenet classifier and general purpose backbone
16
-
17
 
18
  EfficientNetB4 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
19
 
20
- This model is an implementation of EfficientNet-B4 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py).
21
-
22
-
23
- This repository provides scripts to run EfficientNet-B4 on Qualcomm® devices.
24
- More details on model performance across various devices, can be found
25
- [here](https://aihub.qualcomm.com/models/efficientnet_b4).
26
-
27
-
28
-
29
- ### Model Details
30
-
31
- - **Model Type:** Model_use_case.image_classification
32
- - **Model Stats:**
33
- - Model checkpoint: Imagenet
34
- - Input resolution: 380x380
35
- - Number of parameters: 19.3M
36
- - Model size (float): 73.6 MB
37
- - Model size (w8a16): 24.0 MB
38
-
39
- | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
40
- |---|---|---|---|---|---|---|---|---|
41
- | EfficientNet-B4 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 11.94 ms | 0 - 190 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
42
- | EfficientNet-B4 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 11.899 ms | 1 - 156 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
43
- | EfficientNet-B4 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 7.617 ms | 0 - 271 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
44
- | EfficientNet-B4 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 7.752 ms | 1 - 227 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
45
- | EfficientNet-B4 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 3.189 ms | 0 - 2 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
46
- | EfficientNet-B4 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.2 ms | 1 - 2 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
47
- | EfficientNet-B4 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 3.227 ms | 0 - 50 MB | NPU | [EfficientNet-B4.onnx.zip](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx.zip) |
48
- | EfficientNet-B4 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 4.237 ms | 0 - 189 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
49
- | EfficientNet-B4 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 4.225 ms | 1 - 156 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
50
- | EfficientNet-B4 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 2.348 ms | 0 - 251 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
51
- | EfficientNet-B4 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.337 ms | 1 - 214 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
52
- | EfficientNet-B4 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 2.368 ms | 0 - 188 MB | NPU | [EfficientNet-B4.onnx.zip](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx.zip) |
53
- | EfficientNet-B4 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 1.84 ms | 0 - 192 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
54
- | EfficientNet-B4 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 1.845 ms | 1 - 162 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
55
- | EfficientNet-B4 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 1.93 ms | 0 - 135 MB | NPU | [EfficientNet-B4.onnx.zip](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx.zip) |
56
- | EfficientNet-B4 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 1.493 ms | 0 - 192 MB | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
57
- | EfficientNet-B4 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 1.524 ms | 1 - 158 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
58
- | EfficientNet-B4 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 1.671 ms | 0 - 136 MB | NPU | [EfficientNet-B4.onnx.zip](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx.zip) |
59
- | EfficientNet-B4 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 3.489 ms | 1 - 1 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.dlc) |
60
- | EfficientNet-B4 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 3.263 ms | 45 - 45 MB | NPU | [EfficientNet-B4.onnx.zip](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx.zip) |
61
- | EfficientNet-B4 | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | QNN_DLC | 17.887 ms | 0 - 205 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
62
- | EfficientNet-B4 | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 8.645 ms | 0 - 2 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
63
- | EfficientNet-B4 | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 6.564 ms | 0 - 185 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
64
- | EfficientNet-B4 | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 4.115 ms | 0 - 242 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
65
- | EfficientNet-B4 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.408 ms | 0 - 2 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
66
- | EfficientNet-B4 | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 3.838 ms | 0 - 186 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
67
- | EfficientNet-B4 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.276 ms | 0 - 240 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
68
- | EfficientNet-B4 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 1.601 ms | 0 - 193 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
69
- | EfficientNet-B4 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 3.616 ms | 0 - 196 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
70
- | EfficientNet-B4 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 1.316 ms | 0 - 187 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
71
- | EfficientNet-B4 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 3.768 ms | 0 - 0 MB | NPU | [EfficientNet-B4.dlc](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4_w8a16.dlc) |
72
-
73
-
74
-
75
-
76
- ## Installation
77
-
78
-
79
- Install the package via pip:
80
- ```bash
81
- pip install qai-hub-models
82
- ```
83
-
84
-
85
- ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
86
-
87
- Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
88
- Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
89
-
90
- With this API token, you can configure your client to run models on the cloud
91
- hosted devices.
92
- ```bash
93
- qai-hub configure --api_token API_TOKEN
94
- ```
95
- Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
96
-
97
-
98
-
99
- ## Demo off target
100
-
101
- The package contains a simple end-to-end demo that downloads pre-trained
102
- weights and runs this model on a sample input.
103
-
104
- ```bash
105
- python -m qai_hub_models.models.efficientnet_b4.demo
106
- ```
107
-
108
- The above demo runs a reference implementation of pre-processing, model
109
- inference, and post processing.
110
-
111
- **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
112
- environment, please add the following to your cell (instead of the above).
113
- ```
114
- %run -m qai_hub_models.models.efficientnet_b4.demo
115
- ```
116
-
117
-
118
- ### Run model on a cloud-hosted device
119
-
120
- In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
121
- device. This script does the following:
122
- * Performance check on-device on a cloud-hosted device
123
- * Downloads compiled assets that can be deployed on-device for Android.
124
- * Accuracy check between PyTorch and on-device outputs.
125
-
126
- ```bash
127
- python -m qai_hub_models.models.efficientnet_b4.export
128
- ```
129
-
130
-
131
-
132
- ## How does this work?
133
-
134
- This [export script](https://aihub.qualcomm.com/models/efficientnet_b4/qai_hub_models/models/EfficientNet-B4/export.py)
135
- leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
136
- on-device. Lets go through each step below in detail:
137
-
138
- Step 1: **Compile model for on-device deployment**
139
-
140
- To compile a PyTorch model for on-device deployment, we first trace the model
141
- in memory using the `jit.trace` and then call the `submit_compile_job` API.
142
-
143
- ```python
144
- import torch
145
-
146
- import qai_hub as hub
147
- from qai_hub_models.models.efficientnet_b4 import Model
148
-
149
- # Load the model
150
- torch_model = Model.from_pretrained()
151
-
152
- # Device
153
- device = hub.Device("Samsung Galaxy S25")
154
-
155
- # Trace model
156
- input_shape = torch_model.get_input_spec()
157
- sample_inputs = torch_model.sample_inputs()
158
-
159
- pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
160
-
161
- # Compile model on a specific device
162
- compile_job = hub.submit_compile_job(
163
- model=pt_model,
164
- device=device,
165
- input_specs=torch_model.get_input_spec(),
166
- )
167
-
168
- # Get target model to run on-device
169
- target_model = compile_job.get_target_model()
170
-
171
- ```
172
-
173
-
174
- Step 2: **Performance profiling on cloud-hosted device**
175
-
176
- After compiling models from step 1. Models can be profiled model on-device using the
177
- `target_model`. Note that this scripts runs the model on a device automatically
178
- provisioned in the cloud. Once the job is submitted, you can navigate to a
179
- provided job URL to view a variety of on-device performance metrics.
180
- ```python
181
- profile_job = hub.submit_profile_job(
182
- model=target_model,
183
- device=device,
184
- )
185
-
186
- ```
187
-
188
- Step 3: **Verify on-device accuracy**
189
-
190
- To verify the accuracy of the model on-device, you can run on-device inference
191
- on sample input data on the same cloud hosted device.
192
- ```python
193
- input_data = torch_model.sample_inputs()
194
- inference_job = hub.submit_inference_job(
195
- model=target_model,
196
- device=device,
197
- inputs=input_data,
198
- )
199
- on_device_output = inference_job.download_output_data()
200
-
201
- ```
202
- With the output of the model, you can compute like PSNR, relative errors or
203
- spot check the output with expected output.
204
-
205
- **Note**: This on-device profiling and inference requires access to Qualcomm®
206
- AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
207
-
208
-
209
-
210
- ## Run demo on a cloud-hosted device
211
-
212
- You can also run the demo on-device.
213
-
214
- ```bash
215
- python -m qai_hub_models.models.efficientnet_b4.demo --eval-mode on-device
216
- ```
217
-
218
- **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
219
- environment, please add the following to your cell (instead of the above).
220
- ```
221
- %run -m qai_hub_models.models.efficientnet_b4.demo -- --eval-mode on-device
222
- ```
223
-
224
-
225
- ## Deploying compiled model to Android
226
-
227
-
228
- The models can be deployed using multiple runtimes:
229
- - TensorFlow Lite (`.tflite` export): [This
230
- tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
231
- guide to deploy the .tflite model in an Android application.
232
-
233
-
234
- - QNN (`.so` export ): This [sample
235
- app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
236
- provides instructions on how to use the `.so` shared library in an Android application.
237
-
238
-
239
- ## View on Qualcomm® AI Hub
240
- Get more details on EfficientNet-B4's performance across various devices [here](https://aihub.qualcomm.com/models/efficientnet_b4).
241
- Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
242
-
243
 
244
  ## License
245
  * The license for the original implementation of EfficientNet-B4 can be found
246
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).
247
 
248
-
249
-
250
  ## References
251
  * [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://arxiv.org/abs/1905.11946)
252
  * [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py)
253
 
254
-
255
-
256
  ## Community
257
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
258
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
259
-
260
-
 
11
 
12
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/web-assets/model_demo.png)
13
 
14
+ # EfficientNet-B4: Optimized for Qualcomm Devices
 
 
15
 
16
  EfficientNetB4 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
17
 
18
+ This is based on the implementation of EfficientNet-B4 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py).
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/efficientnet_b4) 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
+ There are two ways to deploy this model on your device:
25
+
26
+ ### Option 1: Download Pre-Exported Models
27
+
28
+ Below are pre-exported model assets ready for deployment.
29
+
30
+ | Runtime | Precision | Chipset | SDK Versions | Download |
31
+ |---|---|---|---|---|
32
+ | ONNX | float | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.46.1/efficientnet_b4-onnx-float.zip)
33
+ | 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/efficientnet_b4/releases/v0.46.1/efficientnet_b4-onnx-w8a16.zip)
34
+ | QNN_DLC | float | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.46.1/efficientnet_b4-qnn_dlc-float.zip)
35
+ | QNN_DLC | w8a16 | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.46.1/efficientnet_b4-qnn_dlc-w8a16.zip)
36
+ | 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/efficientnet_b4/releases/v0.46.1/efficientnet_b4-tflite-float.zip)
37
+
38
+ For more device-specific assets and performance metrics, visit **[EfficientNet-B4 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/efficientnet_b4)**.
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/efficientnet_b4) 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 [EfficientNet-B4 on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/efficientnet_b4) for usage instructions.
51
+
52
+ ## Model Details
53
+
54
+ **Model Type:** Model_use_case.image_classification
55
+
56
+ **Model Stats:**
57
+ - Model checkpoint: Imagenet
58
+ - Input resolution: 380x380
59
+ - Number of parameters: 19.3M
60
+ - Model size (float): 73.6 MB
61
+ - Model size (w8a16): 24.0 MB
62
+
63
+ ## Performance Summary
64
+ | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
65
+ |---|---|---|---|---|---|---
66
+ | EfficientNet-B4 | ONNX | float | Snapdragon® X Elite | 3.25 ms | 45 - 45 MB | NPU
67
+ | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.451 ms | 0 - 186 MB | NPU
68
+ | EfficientNet-B4 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 3.185 ms | 0 - 107 MB | NPU
69
+ | EfficientNet-B4 | ONNX | float | Qualcomm® QCS9075 | 4.16 ms | 0 - 4 MB | NPU
70
+ | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.933 ms | 0 - 135 MB | NPU
71
+ | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.656 ms | 0 - 137 MB | NPU
72
+ | EfficientNet-B4 | QNN_DLC | float | Snapdragon® X Elite | 3.598 ms | 1 - 1 MB | NPU
73
+ | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.417 ms | 0 - 124 MB | NPU
74
+ | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 12.05 ms | 1 - 68 MB | NPU
75
+ | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 3.315 ms | 1 - 2 MB | NPU
76
+ | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS9075 | 4.193 ms | 1 - 3 MB | NPU
77
+ | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 7.805 ms | 0 - 145 MB | NPU
78
+ | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.858 ms | 1 - 74 MB | NPU
79
+ | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.507 ms | 0 - 73 MB | NPU
80
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® X Elite | 3.786 ms | 0 - 0 MB | NPU
81
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.328 ms | 0 - 151 MB | NPU
82
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 8.737 ms | 2 - 4 MB | NPU
83
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 6.559 ms | 0 - 99 MB | NPU
84
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 3.429 ms | 0 - 2 MB | NPU
85
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 3.79 ms | 0 - 2 MB | NPU
86
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 17.274 ms | 0 - 229 MB | NPU
87
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 4.203 ms | 0 - 153 MB | NPU
88
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 1.599 ms | 0 - 102 MB | NPU
89
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 3.621 ms | 0 - 107 MB | NPU
90
+ | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.319 ms | 0 - 102 MB | NPU
91
+ | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.421 ms | 0 - 168 MB | NPU
92
+ | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 12.112 ms | 0 - 106 MB | NPU
93
+ | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 3.326 ms | 0 - 3 MB | NPU
94
+ | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS9075 | 4.179 ms | 0 - 48 MB | NPU
95
+ | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 7.901 ms | 0 - 186 MB | NPU
96
+ | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.857 ms | 0 - 111 MB | NPU
97
+ | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.508 ms | 0 - 110 MB | NPU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
98
 
99
  ## License
100
  * The license for the original implementation of EfficientNet-B4 can be found
101
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).
102
 
 
 
103
  ## References
104
  * [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://arxiv.org/abs/1905.11946)
105
  * [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py)
106
 
 
 
107
  ## Community
108
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
109
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
 
 
tool-versions.yaml DELETED
@@ -1,3 +0,0 @@
1
- tool_versions:
2
- qnn_dlc:
3
- qairt: 2.41.0.251128145156_191518