qaihm-bot commited on
Commit
4de5cc3
·
verified ·
1 Parent(s): d140bbf

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +188 -0
README.md ADDED
@@ -0,0 +1,188 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ license: bsd-3-clause
4
+ pipeline_tag: object-detection
5
+ tags:
6
+ - real_time
7
+ - quantized
8
+ - android
9
+
10
+ ---
11
+
12
+ ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/face_det_lite_quantized/web-assets/model_demo.png)
13
+
14
+ # Lightweight-Face-Detection-Quantized: Optimized for Mobile Deployment
15
+ ## face_det_lite_quantized is a face detection model
16
+
17
+
18
+ face_det_lite_quantized is a machine learning model that detect face in the images
19
+
20
+ This model is an implementation of Lightweight-Face-Detection-Quantized found [here](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/face_det_lite/model.py).
21
+
22
+
23
+ This repository provides scripts to run Lightweight-Face-Detection-Quantized on Qualcomm® devices.
24
+ More details on model performance across various devices, can be found
25
+ [here](https://aihub.qualcomm.com/models/face_det_lite_quantized).
26
+
27
+
28
+ ### Model Details
29
+
30
+ - **Model Type:** Object detection
31
+ - **Model Stats:**
32
+ - Model checkpoint: qfd360_sl_model.pt
33
+ - Inference latency: RealTime
34
+ - Input resolution: 480x640
35
+
36
+ | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
+ |---|---|---|---|---|---|---|---|---|
38
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.281 ms | 0 - 8 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
39
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.35 ms | 0 - 20 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.so) |
40
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 0.623 ms | 0 - 11 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.onnx](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.onnx) |
41
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.199 ms | 0 - 15 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
42
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.248 ms | 0 - 18 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.so](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.so) |
43
+ | Lightweight-Face-Detection-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 0.473 ms | 0 - 36 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.onnx](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.onnx) |
44
+ | Lightweight-Face-Detection-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.17 ms | 0 - 11 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
45
+ | Lightweight-Face-Detection-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.242 ms | 0 - 16 MB | INT8 | NPU | Use Export Script |
46
+ | Lightweight-Face-Detection-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 0.478 ms | 0 - 25 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.onnx](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.onnx) |
47
+ | Lightweight-Face-Detection-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 0.802 ms | 0 - 15 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
48
+ | Lightweight-Face-Detection-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 1.012 ms | 0 - 7 MB | INT8 | NPU | Use Export Script |
49
+ | Lightweight-Face-Detection-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 6.076 ms | 0 - 10 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
50
+ | Lightweight-Face-Detection-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.279 ms | 0 - 4 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
51
+ | Lightweight-Face-Detection-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.331 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
52
+ | Lightweight-Face-Detection-Quantized | SA7255P ADP | SA7255P | TFLITE | 3.157 ms | 0 - 16 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
53
+ | Lightweight-Face-Detection-Quantized | SA7255P ADP | SA7255P | QNN | 3.504 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
54
+ | Lightweight-Face-Detection-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.279 ms | 0 - 32 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
55
+ | Lightweight-Face-Detection-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.337 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
56
+ | Lightweight-Face-Detection-Quantized | SA8295P ADP | SA8295P | TFLITE | 0.678 ms | 0 - 10 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
57
+ | Lightweight-Face-Detection-Quantized | SA8295P ADP | SA8295P | QNN | 0.838 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
58
+ | Lightweight-Face-Detection-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.28 ms | 0 - 4 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
59
+ | Lightweight-Face-Detection-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.336 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
60
+ | Lightweight-Face-Detection-Quantized | SA8775P ADP | SA8775P | TFLITE | 0.606 ms | 0 - 16 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
61
+ | Lightweight-Face-Detection-Quantized | SA8775P ADP | SA8775P | QNN | 0.828 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
62
+ | Lightweight-Face-Detection-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.341 ms | 0 - 15 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.tflite](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.tflite) |
63
+ | Lightweight-Face-Detection-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.418 ms | 0 - 17 MB | INT8 | NPU | Use Export Script |
64
+ | Lightweight-Face-Detection-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.436 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
65
+ | Lightweight-Face-Detection-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.698 ms | 2 - 2 MB | INT8 | NPU | [Lightweight-Face-Detection-Quantized.onnx](https://huggingface.co/qualcomm/Lightweight-Face-Detection-Quantized/blob/main/Lightweight-Face-Detection-Quantized.onnx) |
66
+
67
+
68
+
69
+
70
+ ## Installation
71
+
72
+ This model can be installed as a Python package via pip.
73
+
74
+ ```bash
75
+ pip install qai-hub-models
76
+ ```
77
+
78
+
79
+ ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
80
+
81
+ Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
82
+ Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
83
+
84
+ With this API token, you can configure your client to run models on the cloud
85
+ hosted devices.
86
+ ```bash
87
+ qai-hub configure --api_token API_TOKEN
88
+ ```
89
+ Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.
90
+
91
+
92
+
93
+ ## Demo off target
94
+
95
+ The package contains a simple end-to-end demo that downloads pre-trained
96
+ weights and runs this model on a sample input.
97
+
98
+ ```bash
99
+ python -m qai_hub_models.models.face_det_lite_quantized.demo
100
+ ```
101
+
102
+ The above demo runs a reference implementation of pre-processing, model
103
+ inference, and post processing.
104
+
105
+ **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
106
+ environment, please add the following to your cell (instead of the above).
107
+ ```
108
+ %run -m qai_hub_models.models.face_det_lite_quantized.demo
109
+ ```
110
+
111
+
112
+ ### Run model on a cloud-hosted device
113
+
114
+ In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
115
+ device. This script does the following:
116
+ * Performance check on-device on a cloud-hosted device
117
+ * Downloads compiled assets that can be deployed on-device for Android.
118
+ * Accuracy check between PyTorch and on-device outputs.
119
+
120
+ ```bash
121
+ python -m qai_hub_models.models.face_det_lite_quantized.export
122
+ ```
123
+ ```
124
+ Profiling Results
125
+ ------------------------------------------------------------
126
+ Lightweight-Face-Detection-Quantized
127
+ Device : Samsung Galaxy S23 (13)
128
+ Runtime : TFLITE
129
+ Estimated inference time (ms) : 0.3
130
+ Estimated peak memory usage (MB): [0, 8]
131
+ Total # Ops : 90
132
+ Compute Unit(s) : NPU (90 ops)
133
+ ```
134
+
135
+
136
+
137
+
138
+ ## Run demo on a cloud-hosted device
139
+
140
+ You can also run the demo on-device.
141
+
142
+ ```bash
143
+ python -m qai_hub_models.models.face_det_lite_quantized.demo --on-device
144
+ ```
145
+
146
+ **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
147
+ environment, please add the following to your cell (instead of the above).
148
+ ```
149
+ %run -m qai_hub_models.models.face_det_lite_quantized.demo -- --on-device
150
+ ```
151
+
152
+
153
+ ## Deploying compiled model to Android
154
+
155
+
156
+ The models can be deployed using multiple runtimes:
157
+ - TensorFlow Lite (`.tflite` export): [This
158
+ tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
159
+ guide to deploy the .tflite model in an Android application.
160
+
161
+
162
+ - QNN (`.so` export ): This [sample
163
+ app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
164
+ provides instructions on how to use the `.so` shared library in an Android application.
165
+
166
+
167
+ ## View on Qualcomm® AI Hub
168
+ Get more details on Lightweight-Face-Detection-Quantized's performance across various devices [here](https://aihub.qualcomm.com/models/face_det_lite_quantized).
169
+ Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
170
+
171
+
172
+ ## License
173
+ * The license for the original implementation of Lightweight-Face-Detection-Quantized can be found [here](https://github.com/qcom-ai-hub/ai-hub-models-internal/blob/main/LICENSE).
174
+ * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
175
+
176
+
177
+
178
+ ## References
179
+ * [None](None)
180
+ * [Source Model Implementation](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/face_det_lite/model.py)
181
+
182
+
183
+
184
+ ## Community
185
+ * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
186
+ * For questions or feedback please [reach out to us](mailto:[email protected]).
187
+
188
+