Hugging Face's logo

qualcomm
/
3D-Deep-BOX

Object Detection
PyTorch
real_time
android
Model card Files
xet
 Community
3D-Deep-BOX
File size: 20,215 Bytes
54faf13
 
e2daa10
54faf13
 
 
 
 
 
 
 
 
5cec659
54faf13
 
 
5cec659
cbf62a7
5cec659
 
 
 
 
 
 
 
 
 
 
 
cbf62a7
 
 
 
 
5cec659
 
 
 
 
 
cbf62a7
5cec659
 
 
 
 
 
cbf62a7
5cec659
 
 
 
 
 
cbf62a7
 
5cec659
cbf62a7
 
c8745b5
cbf62a7
 
 
5cec659
 
 
 
cbf62a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25fc0a6
cbf62a7
 
25fc0a6
cbf62a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
54faf13
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: object-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/web-assets/model_demo.png)

# 3D-Deep-BOX: Optimized for Qualcomm Devices

3D Deep Box is a machine learning model that predicts 3D bounding boxes and classes of objects in an image.

This is based on the implementation of 3D-Deep-BOX found [here](https://github.com/skhadem/3D-BoundingBox/).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/deepbox) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

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.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/releases/v0.60.0/deepbox-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/releases/v0.60.0/deepbox-onnx-w8a16.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/releases/v0.60.0/deepbox-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/releases/v0.60.0/deepbox-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deepbox/releases/v0.60.0/deepbox-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[3D-Deep-BOX on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/deepbox)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/deepbox) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [3D-Deep-BOX on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/deepbox) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- 2D detector: COCO-pretrained YOLOv3 (paper uses KITTI-trained MS-CNN)
- Accuracy (2D AP E/M/H): ~73/59/45 (COCO YOLOv3) vs 92.98/89.04/77.17 (paper MS-CNN). Gap is due to the 2D detector only; VGG 3D head weights are identical to the paper.
- Input resolution(VGG): 224x224
- Input resolution(YOLO): 416x416
- Model checkpoint: YOLOv3
- Model size (vgg_3d_detection) (float): 176 MB
- Model size (yolo_2d_detection) (float): 236 MB
- Number of parameters (vgg_3d_detection): 46.1M
- Number of parameters (yolo_2d_detection): 61.9M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| vgg_3d_detection | ONNX | float | Snapdragon® X2 Elite | 2.607 ms | 2 - 2 MB | NPU
| vgg_3d_detection | ONNX | float | Snapdragon® X Elite | 4.877 ms | 88 - 88 MB | NPU
| vgg_3d_detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.772 ms | 0 - 273 MB | NPU
| vgg_3d_detection | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 9.6 ms | 0 - 207 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 10.017 ms | 0 - 4 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.896 ms | 0 - 110 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® QCS8450 | 9.6 ms | 0 - 207 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 9.06 ms | 1 - 4 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 4.877 ms | 88 - 88 MB | NPU
| vgg_3d_detection | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 3.011 ms | 0 - 200 MB | NPU
| vgg_3d_detection | ONNX | float | Snapdragon® 8 Elite Mobile | 3.011 ms | 0 - 200 MB | NPU
| vgg_3d_detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.471 ms | 0 - 200 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® X2 Elite | 1.633 ms | 1 - 1 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® X Elite | 2.703 ms | 45 - 45 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.076 ms | 0 - 97 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 3.522 ms | 0 - 97 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 351.406 ms | 0 - 3 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 3.293 ms | 0 - 4 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.683 ms | 0 - 50 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® QCS8450 | 3.522 ms | 0 - 97 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 2.799 ms | 0 - 3 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 2.703 ms | 45 - 45 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 1877.733 ms | 1 - 199 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 5.168 ms | 0 - 197 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 1.698 ms | 0 - 58 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 1.698 ms | 0 - 58 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.406 ms | 0 - 57 MB | NPU
| vgg_3d_detection | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 5.168 ms | 0 - 197 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® X2 Elite | 2.807 ms | 1 - 1 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® X Elite | 5.328 ms | 1 - 1 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.904 ms | 0 - 274 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 9.999 ms | 0 - 205 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 10.172 ms | 1 - 4 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 33.444 ms | 1 - 198 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.153 ms | 1 - 2 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® SA8775P | 9.239 ms | 1 - 200 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® SA8650P | 9.239 ms | 1 - 200 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® SA8255P | 9.239 ms | 1 - 200 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® QCS8450 | 9.999 ms | 0 - 205 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 9.039 ms | 1 - 3 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.328 ms | 1 - 1 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 3.154 ms | 0 - 199 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® SA7255P | 33.444 ms | 1 - 198 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Qualcomm® SA8295P | 10.559 ms | 1 - 195 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 3.154 ms | 0 - 199 MB | NPU
| vgg_3d_detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.517 ms | 1 - 200 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 1.75 ms | 0 - 0 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® X Elite | 2.868 ms | 0 - 0 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.108 ms | 1 - 98 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 3.516 ms | 0 - 96 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 351.42 ms | 0 - 2 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 3.289 ms | 0 - 3 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 9.538 ms | 0 - 54 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.729 ms | 0 - 2 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® SA8775P | 2.953 ms | 0 - 57 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® SA8650P | 2.953 ms | 0 - 57 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® SA8255P | 2.953 ms | 0 - 57 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 3.516 ms | 0 - 96 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 2.819 ms | 2 - 4 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 2.868 ms | 0 - 0 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 5.051 ms | 0 - 195 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 1.702 ms | 0 - 53 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® SA7255P | 9.538 ms | 0 - 54 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Qualcomm® SA8295P | 5.292 ms | 0 - 51 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 1.702 ms | 0 - 53 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.413 ms | 0 - 56 MB | NPU
| vgg_3d_detection | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 5.051 ms | 0 - 195 MB | NPU
| vgg_3d_detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 3.894 ms | 0 - 317 MB | NPU
| vgg_3d_detection | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 10.005 ms | 0 - 250 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 10.181 ms | 0 - 91 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 33.484 ms | 0 - 199 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.139 ms | 0 - 2 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® SA8775P | 9.263 ms | 0 - 201 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® SA8650P | 9.263 ms | 0 - 201 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® SA8255P | 9.263 ms | 0 - 201 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® QCS8450 | 10.005 ms | 0 - 250 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 9.06 ms | 0 - 91 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 3.118 ms | 0 - 202 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® SA7255P | 33.484 ms | 0 - 199 MB | NPU
| vgg_3d_detection | TFLITE | float | Qualcomm® SA8295P | 10.578 ms | 0 - 194 MB | NPU
| vgg_3d_detection | TFLITE | float | Snapdragon® 8 Elite Mobile | 3.118 ms | 0 - 202 MB | NPU
| vgg_3d_detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.518 ms | 0 - 203 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® X2 Elite | 10.28 ms | 2 - 2 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® X Elite | 18.879 ms | 199 - 199 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 14.693 ms | 1 - 233 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 38.419 ms | 2 - 265 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 35.394 ms | 2 - 7 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 19.171 ms | 0 - 225 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® QCS8450 | 38.419 ms | 2 - 265 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 30.546 ms | 2 - 6 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 18.879 ms | 199 - 199 MB | NPU
| yolo_2d_detection | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 11.623 ms | 2 - 193 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® 8 Elite Mobile | 11.623 ms | 2 - 193 MB | NPU
| yolo_2d_detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.501 ms | 2 - 199 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® X2 Elite | 5.773 ms | 2 - 2 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® X Elite | 12.146 ms | 101 - 101 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 8.107 ms | 0 - 430 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 18.97 ms | 0 - 405 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 58.027 ms | 0 - 4 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 11.913 ms | 1 - 5 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 11.895 ms | 0 - 129 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® QCS8450 | 18.97 ms | 0 - 405 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 10.752 ms | 0 - 4 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 12.146 ms | 101 - 101 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 208.66 ms | 0 - 386 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 20.197 ms | 1 - 361 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 6.511 ms | 1 - 237 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 6.511 ms | 1 - 237 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.543 ms | 0 - 247 MB | NPU
| yolo_2d_detection | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 20.197 ms | 1 - 361 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® X2 Elite | 10.443 ms | 2 - 2 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® X Elite | 17.954 ms | 2 - 2 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 13.7 ms | 1 - 226 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 38.21 ms | 0 - 243 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 34.23 ms | 2 - 6 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 117.953 ms | 2 - 188 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 17.565 ms | 2 - 4 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® SA8775P | 28.67 ms | 2 - 189 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® SA8650P | 28.67 ms | 2 - 189 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® SA8255P | 28.67 ms | 2 - 189 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® QCS8450 | 38.21 ms | 0 - 243 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 29.191 ms | 2 - 6 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 17.954 ms | 2 - 2 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 10.846 ms | 1 - 186 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® SA7255P | 117.953 ms | 2 - 188 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Qualcomm® SA8295P | 32.967 ms | 2 - 210 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 10.846 ms | 1 - 186 MB | NPU
| yolo_2d_detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.252 ms | 2 - 196 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 5.959 ms | 1 - 1 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® X Elite | 12.227 ms | 1 - 1 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 7.948 ms | 1 - 420 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 19.566 ms | 1 - 389 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 56.736 ms | 2 - 5 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 11.67 ms | 1 - 4 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 35.674 ms | 1 - 230 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 10.765 ms | 1 - 3 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® SA8775P | 10.733 ms | 1 - 231 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® SA8650P | 10.733 ms | 1 - 231 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® SA8255P | 10.733 ms | 1 - 231 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 19.566 ms | 1 - 389 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 10.449 ms | 1 - 4 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 12.227 ms | 1 - 1 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 20.421 ms | 1 - 353 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 6.444 ms | 0 - 223 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® SA7255P | 35.674 ms | 1 - 230 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Qualcomm® SA8295P | 18.77 ms | 1 - 230 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 6.444 ms | 0 - 223 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.412 ms | 1 - 235 MB | NPU
| yolo_2d_detection | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 20.421 ms | 1 - 353 MB | NPU
| yolo_2d_detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 13.417 ms | 0 - 419 MB | NPU
| yolo_2d_detection | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 39.281 ms | 0 - 434 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 33.547 ms | 0 - 205 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 117.547 ms | 0 - 196 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 16.759 ms | 0 - 2 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® SA8775P | 28.493 ms | 0 - 196 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® SA8650P | 28.493 ms | 0 - 196 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® SA8255P | 28.493 ms | 0 - 196 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® QCS8450 | 39.281 ms | 0 - 434 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 28.739 ms | 0 - 204 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 11.068 ms | 0 - 197 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® SA7255P | 117.547 ms | 0 - 196 MB | NPU
| yolo_2d_detection | TFLITE | float | Qualcomm® SA8295P | 32.207 ms | 0 - 212 MB | NPU
| yolo_2d_detection | TFLITE | float | Snapdragon® 8 Elite Mobile | 11.068 ms | 0 - 197 MB | NPU
| yolo_2d_detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.262 ms | 0 - 202 MB | NPU

## License
* The license for the original implementation of 3D-Deep-BOX can be found
  [here](https://github.com/skhadem/3D-BoundingBox/blob/master/LICENSE).

## References
* [3D Bounding Box Estimation Using Deep Learning and Geometry](https://arxiv.org/abs/1612.00496)
* [Source Model Implementation](https://github.com/skhadem/3D-BoundingBox/)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:[email protected]).