Hugging Face's logo

qualcomm
/
3D-Deep-BOX

Object Detection
PyTorch
LiteRT
real_time
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
4fdc0dd
·
verified ·
1 Parent(s): 18893d5

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +82 -110
README.md CHANGED
@@ -38,64 +38,64 @@ More details on model performance across various devices, can be found
38
 
39
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
40
  |---|---|---|---|---|---|---|---|---|
41
- | Yolo | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 21.944 ms | 0 - 135 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
42
- | Yolo | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 2.435 ms | 2 - 4 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.so) |
43
- | Yolo | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 5.173 ms | 3 - 54 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.onnx) |
44
- | Yolo | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 16.556 ms | 0 - 61 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
45
- | Yolo | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 1.748 ms | 2 - 22 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.so) |
46
- | Yolo | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.528 ms | 1 - 26 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.onnx) |
47
- | Yolo | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 14.082 ms | 0 - 33 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
48
- | Yolo | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 2.087 ms | 2 - 19 MB | FP16 | NPU | Use Export Script |
49
- | Yolo | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 4.32 ms | 2 - 15 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.onnx) |
50
- | Yolo | SA7255P ADP | SA7255P | TFLITE | 67.784 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
51
- | Yolo | SA7255P ADP | SA7255P | QNN | 34.599 ms | 2 - 11 MB | FP16 | NPU | Use Export Script |
52
- | Yolo | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 22.109 ms | 0 - 141 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
53
- | Yolo | SA8255 (Proxy) | SA8255P Proxy | QNN | 2.436 ms | 2 - 4 MB | FP16 | NPU | Use Export Script |
54
- | Yolo | SA8295P ADP | SA8295P | TFLITE | 23.94 ms | 0 - 31 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
55
- | Yolo | SA8295P ADP | SA8295P | QNN | 3.475 ms | 0 - 18 MB | FP16 | NPU | Use Export Script |
56
- | Yolo | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 21.84 ms | 0 - 140 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
57
- | Yolo | SA8650 (Proxy) | SA8650P Proxy | QNN | 2.426 ms | 2 - 4 MB | FP16 | NPU | Use Export Script |
58
- | Yolo | SA8775P ADP | SA8775P | TFLITE | 27.669 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
59
- | Yolo | SA8775P ADP | SA8775P | QNN | 3.837 ms | 2 - 11 MB | FP16 | NPU | Use Export Script |
60
- | Yolo | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 67.784 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
61
- | Yolo | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 34.599 ms | 2 - 11 MB | FP16 | NPU | Use Export Script |
62
- | Yolo | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 21.809 ms | 0 - 136 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
63
- | Yolo | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 2.44 ms | 2 - 5 MB | FP16 | NPU | Use Export Script |
64
- | Yolo | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 27.669 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
65
- | Yolo | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 3.837 ms | 2 - 11 MB | FP16 | NPU | Use Export Script |
66
- | Yolo | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 22.505 ms | 0 - 60 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.tflite) |
67
- | Yolo | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 3.531 ms | 2 - 22 MB | FP16 | NPU | Use Export Script |
68
- | Yolo | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 2.653 ms | 2 - 2 MB | FP16 | NPU | Use Export Script |
69
- | Yolo | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.558 ms | 11 - 11 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo.onnx) |
70
- | VGG | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 4.72 ms | 0 - 667 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
71
- | VGG | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 4.775 ms | 1 - 2 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.so) |
72
- | VGG | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 5.323 ms | 0 - 511 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.onnx) |
73
- | VGG | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 3.572 ms | 0 - 125 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
74
- | VGG | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 3.753 ms | 0 - 18 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.so) |
75
- | VGG | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 4.057 ms | 0 - 82 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.onnx) |
76
- | VGG | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 3.454 ms | 0 - 78 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
77
- | VGG | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 3.427 ms | 1 - 76 MB | FP16 | NPU | Use Export Script |
78
- | VGG | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.811 ms | 1 - 75 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.onnx) |
79
- | VGG | SA7255P ADP | SA7255P | TFLITE | 257.893 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
80
- | VGG | SA7255P ADP | SA7255P | QNN | 257.859 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
81
- | VGG | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 4.721 ms | 0 - 669 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
82
- | VGG | SA8255 (Proxy) | SA8255P Proxy | QNN | 4.801 ms | 1 - 3 MB | FP16 | NPU | Use Export Script |
83
- | VGG | SA8295P ADP | SA8295P | TFLITE | 9.753 ms | 0 - 76 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
84
- | VGG | SA8295P ADP | SA8295P | QNN | 9.913 ms | 1 - 19 MB | FP16 | NPU | Use Export Script |
85
- | VGG | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 4.729 ms | 0 - 664 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
86
- | VGG | SA8650 (Proxy) | SA8650P Proxy | QNN | 4.776 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
87
- | VGG | SA8775P ADP | SA8775P | TFLITE | 10.811 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
88
- | VGG | SA8775P ADP | SA8775P | QNN | 10.713 ms | 0 - 10 MB | FP16 | NPU | Use Export Script |
89
- | VGG | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 257.893 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
90
- | VGG | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 257.859 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
91
- | VGG | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 4.721 ms | 0 - 668 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
92
- | VGG | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 4.772 ms | 1 - 4 MB | FP16 | NPU | Use Export Script |
93
- | VGG | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 10.811 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
94
- | VGG | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 10.713 ms | 0 - 10 MB | FP16 | NPU | Use Export Script |
95
- | VGG | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 8.44 ms | 0 - 125 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.tflite) |
96
- | VGG | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 8.649 ms | 1 - 83 MB | FP16 | NPU | Use Export Script |
97
- | VGG | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 5.023 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
98
- | VGG | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.251 ms | 89 - 89 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG.onnx) |
99
 
100
 
101
 
@@ -156,20 +156,20 @@ python -m qai_hub_models.models.deepbox.export
156
  ```
157
  Profiling Results
158
  ------------------------------------------------------------
159
- Yolo
160
  Device : Samsung Galaxy S23 (13)
161
  Runtime : TFLITE
162
- Estimated inference time (ms) : 21.9
163
- Estimated peak memory usage (MB): [0, 135]
164
- Total # Ops : 128
165
- Compute Unit(s) : NPU (128 ops)
166
 
167
  ------------------------------------------------------------
168
- VGG
169
  Device : Samsung Galaxy S23 (13)
170
  Runtime : TFLITE
171
  Estimated inference time (ms) : 4.7
172
- Estimated peak memory usage (MB): [0, 667]
173
  Total # Ops : 40
174
  Compute Unit(s) : NPU (40 ops)
175
  ```
@@ -193,43 +193,26 @@ import qai_hub as hub
193
  from qai_hub_models.models.deepbox import Model
194
 
195
  # Load the model
196
- model = Model.from_pretrained()
197
- bbox2D_dectector_model = model.bbox2D_dectector
198
- bbox3D_dectector_model = model.bbox3D_dectector
199
 
200
  # Device
201
- device = hub.Device("Samsung Galaxy S23")
202
 
203
  # Trace model
204
- bbox2D_dectector_input_shape = bbox2D_dectector_model.get_input_spec()
205
- bbox2D_dectector_sample_inputs = bbox2D_dectector_model.sample_inputs()
206
 
207
- traced_bbox2D_dectector_model = torch.jit.trace(bbox2D_dectector_model, [torch.tensor(data[0]) for _, data in bbox2D_dectector_sample_inputs.items()])
208
 
209
  # Compile model on a specific device
210
- bbox2D_dectector_compile_job = hub.submit_compile_job(
211
- model=traced_bbox2D_dectector_model ,
212
  device=device,
213
- input_specs=bbox2D_dectector_model.get_input_spec(),
214
  )
215
 
216
  # Get target model to run on-device
217
- bbox2D_dectector_target_model = bbox2D_dectector_compile_job.get_target_model()
218
- # Trace model
219
- bbox3D_dectector_input_shape = bbox3D_dectector_model.get_input_spec()
220
- bbox3D_dectector_sample_inputs = bbox3D_dectector_model.sample_inputs()
221
-
222
- traced_bbox3D_dectector_model = torch.jit.trace(bbox3D_dectector_model, [torch.tensor(data[0]) for _, data in bbox3D_dectector_sample_inputs.items()])
223
-
224
- # Compile model on a specific device
225
- bbox3D_dectector_compile_job = hub.submit_compile_job(
226
- model=traced_bbox3D_dectector_model ,
227
- device=device,
228
- input_specs=bbox3D_dectector_model.get_input_spec(),
229
- )
230
-
231
- # Get target model to run on-device
232
- bbox3D_dectector_target_model = bbox3D_dectector_compile_job.get_target_model()
233
 
234
  ```
235
 
@@ -241,15 +224,11 @@ After compiling models from step 1. Models can be profiled model on-device using
241
  provisioned in the cloud. Once the job is submitted, you can navigate to a
242
  provided job URL to view a variety of on-device performance metrics.
243
  ```python
244
- bbox2D_dectector_profile_job = hub.submit_profile_job(
245
- model=bbox2D_dectector_target_model,
246
- device=device,
247
- )
248
- bbox3D_dectector_profile_job = hub.submit_profile_job(
249
- model=bbox3D_dectector_target_model,
250
  device=device,
251
  )
252
-
253
  ```
254
 
255
  Step 3: **Verify on-device accuracy**
@@ -257,20 +236,13 @@ Step 3: **Verify on-device accuracy**
257
  To verify the accuracy of the model on-device, you can run on-device inference
258
  on sample input data on the same cloud hosted device.
259
  ```python
260
- bbox2D_dectector_input_data = bbox2D_dectector_model.sample_inputs()
261
- bbox2D_dectector_inference_job = hub.submit_inference_job(
262
- model=bbox2D_dectector_target_model,
263
- device=device,
264
- inputs=bbox2D_dectector_input_data,
265
- )
266
- bbox2D_dectector_inference_job.download_output_data()
267
- bbox3D_dectector_input_data = bbox3D_dectector_model.sample_inputs()
268
- bbox3D_dectector_inference_job = hub.submit_inference_job(
269
- model=bbox3D_dectector_target_model,
270
  device=device,
271
- inputs=bbox3D_dectector_input_data,
272
  )
273
- bbox3D_dectector_inference_job.download_output_data()
274
 
275
  ```
276
  With the output of the model, you can compute like PSNR, relative errors or
 
38
 
39
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
40
  |---|---|---|---|---|---|---|---|---|
41
+ | Yolo2DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 22.84 ms | 0 - 136 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
42
+ | Yolo2DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 2.437 ms | 2 - 4 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.so) |
43
+ | Yolo2DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 4.822 ms | 3 - 57 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.onnx) |
44
+ | Yolo2DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 16.285 ms | 0 - 58 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
45
+ | Yolo2DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 1.755 ms | 0 - 18 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.so) |
46
+ | Yolo2DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.318 ms | 0 - 27 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.onnx) |
47
+ | Yolo2DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 17.442 ms | 0 - 33 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
48
+ | Yolo2DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 1.653 ms | 2 - 19 MB | FP16 | NPU | Use Export Script |
49
+ | Yolo2DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 4.132 ms | 2 - 17 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.onnx) |
50
+ | Yolo2DDetection | SA7255P ADP | SA7255P | TFLITE | 67.55 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
51
+ | Yolo2DDetection | SA7255P ADP | SA7255P | QNN | 34.614 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
52
+ | Yolo2DDetection | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 22.35 ms | 0 - 129 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
53
+ | Yolo2DDetection | SA8255 (Proxy) | SA8255P Proxy | QNN | 2.435 ms | 2 - 5 MB | FP16 | NPU | Use Export Script |
54
+ | Yolo2DDetection | SA8295P ADP | SA8295P | TFLITE | 24.002 ms | 0 - 31 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
55
+ | Yolo2DDetection | SA8295P ADP | SA8295P | QNN | 3.484 ms | 0 - 18 MB | FP16 | NPU | Use Export Script |
56
+ | Yolo2DDetection | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 22.3 ms | 0 - 141 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
57
+ | Yolo2DDetection | SA8650 (Proxy) | SA8650P Proxy | QNN | 2.44 ms | 2 - 12 MB | FP16 | NPU | Use Export Script |
58
+ | Yolo2DDetection | SA8775P ADP | SA8775P | TFLITE | 28.223 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
59
+ | Yolo2DDetection | SA8775P ADP | SA8775P | QNN | 3.868 ms | 2 - 12 MB | FP16 | NPU | Use Export Script |
60
+ | Yolo2DDetection | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 67.55 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
61
+ | Yolo2DDetection | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 34.614 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
62
+ | Yolo2DDetection | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 22.221 ms | 0 - 126 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
63
+ | Yolo2DDetection | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 2.435 ms | 2 - 4 MB | FP16 | NPU | Use Export Script |
64
+ | Yolo2DDetection | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 28.223 ms | 0 - 29 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
65
+ | Yolo2DDetection | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 3.868 ms | 2 - 12 MB | FP16 | NPU | Use Export Script |
66
+ | Yolo2DDetection | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 22.239 ms | 0 - 59 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.tflite) |
67
+ | Yolo2DDetection | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 3.322 ms | 2 - 23 MB | FP16 | NPU | Use Export Script |
68
+ | Yolo2DDetection | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 2.637 ms | 2 - 2 MB | FP16 | NPU | Use Export Script |
69
+ | Yolo2DDetection | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.49 ms | 3 - 3 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/Yolo2DDetection.onnx) |
70
+ | VGG3DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 4.728 ms | 0 - 674 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
71
+ | VGG3DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 4.777 ms | 1 - 3 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.so) |
72
+ | VGG3DDetection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 5.127 ms | 0 - 451 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.onnx) |
73
+ | VGG3DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 3.555 ms | 0 - 125 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
74
+ | VGG3DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 3.754 ms | 0 - 18 MB | FP16 | NPU | [3D-Deep-BOX.so](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.so) |
75
+ | VGG3DDetection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.956 ms | 0 - 83 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.onnx) |
76
+ | VGG3DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 3.451 ms | 0 - 79 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
77
+ | VGG3DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 3.43 ms | 1 - 76 MB | FP16 | NPU | Use Export Script |
78
+ | VGG3DDetection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.792 ms | 1 - 80 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.onnx) |
79
+ | VGG3DDetection | SA7255P ADP | SA7255P | TFLITE | 257.868 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
80
+ | VGG3DDetection | SA7255P ADP | SA7255P | QNN | 257.837 ms | 1 - 11 MB | FP16 | NPU | Use Export Script |
81
+ | VGG3DDetection | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 4.727 ms | 0 - 663 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
82
+ | VGG3DDetection | SA8255 (Proxy) | SA8255P Proxy | QNN | 4.782 ms | 1 - 3 MB | FP16 | NPU | Use Export Script |
83
+ | VGG3DDetection | SA8295P ADP | SA8295P | TFLITE | 9.744 ms | 0 - 76 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
84
+ | VGG3DDetection | SA8295P ADP | SA8295P | QNN | 9.927 ms | 1 - 18 MB | FP16 | NPU | Use Export Script |
85
+ | VGG3DDetection | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 4.732 ms | 0 - 666 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
86
+ | VGG3DDetection | SA8650 (Proxy) | SA8650P Proxy | QNN | 4.785 ms | 1 - 3 MB | FP16 | NPU | Use Export Script |
87
+ | VGG3DDetection | SA8775P ADP | SA8775P | TFLITE | 10.781 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
88
+ | VGG3DDetection | SA8775P ADP | SA8775P | QNN | 10.68 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
89
+ | VGG3DDetection | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 257.868 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
90
+ | VGG3DDetection | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 257.837 ms | 1 - 11 MB | FP16 | NPU | Use Export Script |
91
+ | VGG3DDetection | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 4.722 ms | 0 - 666 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
92
+ | VGG3DDetection | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 4.777 ms | 1 - 3 MB | FP16 | NPU | Use Export Script |
93
+ | VGG3DDetection | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 10.781 ms | 0 - 73 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
94
+ | VGG3DDetection | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 10.68 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
95
+ | VGG3DDetection | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 8.457 ms | 0 - 125 MB | FP16 | NPU | [3D-Deep-BOX.tflite](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.tflite) |
96
+ | VGG3DDetection | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 8.718 ms | 1 - 83 MB | FP16 | NPU | Use Export Script |
97
+ | VGG3DDetection | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 5.011 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
98
+ | VGG3DDetection | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 4.982 ms | 89 - 89 MB | FP16 | NPU | [3D-Deep-BOX.onnx](https://huggingface.co/qualcomm/3D-Deep-BOX/blob/main/VGG3DDetection.onnx) |
99
 
100
 
101
 
 
156
  ```
157
  Profiling Results
158
  ------------------------------------------------------------
159
+ Yolo2DDetection
160
  Device : Samsung Galaxy S23 (13)
161
  Runtime : TFLITE
162
+ Estimated inference time (ms) : 22.8
163
+ Estimated peak memory usage (MB): [0, 136]
164
+ Total # Ops : 129
165
+ Compute Unit(s) : NPU (129 ops)
166
 
167
  ------------------------------------------------------------
168
+ VGG3DDetection
169
  Device : Samsung Galaxy S23 (13)
170
  Runtime : TFLITE
171
  Estimated inference time (ms) : 4.7
172
+ Estimated peak memory usage (MB): [0, 674]
173
  Total # Ops : 40
174
  Compute Unit(s) : NPU (40 ops)
175
  ```
 
193
  from qai_hub_models.models.deepbox import Model
194
 
195
  # Load the model
196
+ torch_model = Model.from_pretrained()
 
 
197
 
198
  # Device
199
+ device = hub.Device("Samsung Galaxy S24")
200
 
201
  # Trace model
202
+ input_shape = torch_model.get_input_spec()
203
+ sample_inputs = torch_model.sample_inputs()
204
 
205
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
206
 
207
  # Compile model on a specific device
208
+ compile_job = hub.submit_compile_job(
209
+ model=pt_model,
210
  device=device,
211
+ input_specs=torch_model.get_input_spec(),
212
  )
213
 
214
  # Get target model to run on-device
215
+ target_model = compile_job.get_target_model()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
216
 
217
  ```
218
 
 
224
  provisioned in the cloud. Once the job is submitted, you can navigate to a
225
  provided job URL to view a variety of on-device performance metrics.
226
  ```python
227
+ profile_job = hub.submit_profile_job(
228
+ model=target_model,
 
 
 
 
229
  device=device,
230
  )
231
+
232
  ```
233
 
234
  Step 3: **Verify on-device accuracy**
 
236
  To verify the accuracy of the model on-device, you can run on-device inference
237
  on sample input data on the same cloud hosted device.
238
  ```python
239
+ input_data = torch_model.sample_inputs()
240
+ inference_job = hub.submit_inference_job(
241
+ model=target_model,
 
 
 
 
 
 
 
242
  device=device,
243
+ inputs=input_data,
244
  )
245
+ on_device_output = inference_job.download_output_data()
246
 
247
  ```
248
  With the output of the model, you can compute like PSNR, relative errors or