Hugging Face's logo

qualcomm
/
RTMDet

Object Detection
PyTorch
real_time
android
Model card Files
xet
 Community
qaihm-bot commited on
Commit
8fa355e
·
verified ·
1 Parent(s): 0464784

v0.58.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.58.0 for changelog.

Files changed (1) hide show
  1. README.md +43 -43
README.md CHANGED
@@ -15,18 +15,18 @@ pipeline_tag: object-detection
15
  RTMDet is a highly efficient model for real-time object detection,capable of predicting both the bounding boxes and classes of objects within an image.It is highly optimized for real-time applications, making it reliable for industrial and commercial use
16
 
17
  This is based on the implementation of RTMDet found [here](https://github.com/open-mmlab/mmdetection/tree/3.x/configs/rtmdet).
18
- 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/v0.57.3/src/qai_hub_models/models/rtmdet) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  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.
21
 
22
  ## Getting Started
23
  Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
24
- Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.57.3/src/qai_hub_models/models/rtmdet) Python library to compile and export the model with your own:
25
  - Custom weights (e.g., fine-tuned checkpoints)
26
  - Custom input shapes
27
  - Target device and runtime configurations
28
 
29
- See our repository for [RTMDet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.57.3/src/qai_hub_models/models/rtmdet) for usage instructions.
30
 
31
 
32
  ## Model Details
@@ -42,46 +42,46 @@ See our repository for [RTMDet on GitHub](https://github.com/qualcomm/ai-hub-mod
42
  ## Performance Summary
43
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
44
  |---|---|---|---|---|---|---
45
- | RTMDet | ONNX | float | Snapdragon® X2 Elite | 7.996 ms | 208 - 208 MB | NPU
46
- | RTMDet | ONNX | float | Snapdragon® X Elite | 14.426 ms | 177 - 177 MB | NPU
47
- | RTMDet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 10.673 ms | 5 - 220 MB | NPU
48
- | RTMDet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 35.616 ms | 5 - 282 MB | NPU
49
- | RTMDet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 13.933 ms | 0 - 58 MB | NPU
50
- | RTMDet | ONNX | float | Qualcomm® QCS8450 | 35.616 ms | 5 - 282 MB | NPU
51
- | RTMDet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.847 ms | 3 - 182 MB | NPU
52
- | RTMDet | ONNX | float | Qualcomm® QCS9075 | 23.452 ms | 5 - 50 MB | NPU
53
- | RTMDet | ONNX | float | Snapdragon® 8 Elite Mobile | 8.806 ms | 3 - 185 MB | NPU
54
- | RTMDet | ONNX | float | Qualcomm® QCS8750 | 8.806 ms | 3 - 185 MB | NPU
55
- | RTMDet | ONNX | float | Qualcomm® QCS7181 | 14.426 ms | 177 - 177 MB | NPU
56
- | RTMDet | ONNX | w8a16 | Snapdragon® X2 Elite | 5.022 ms | 210 - 210 MB | NPU
57
- | RTMDet | ONNX | w8a16 | Snapdragon® X Elite | 12.043 ms | 147 - 147 MB | NPU
58
- | RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 7.514 ms | 3 - 331 MB | NPU
59
- | RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 16.991 ms | 3 - 333 MB | NPU
60
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS6490 | 42.171 ms | 2 - 48 MB | NPU
61
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 11.495 ms | 2 - 6 MB | NPU
62
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS8450 | 16.991 ms | 3 - 333 MB | NPU
63
- | RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.223 ms | 1 - 299 MB | NPU
64
- | RTMDet | ONNX | w8a16 | Qualcomm® QCM6690 | 162.895 ms | 3 - 322 MB | NPU
65
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS9075 | 14.297 ms | 2 - 48 MB | NPU
66
- | RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 5.541 ms | 1 - 292 MB | NPU
67
- | RTMDet | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 20.349 ms | 3 - 298 MB | NPU
68
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS7790 | 20.349 ms | 3 - 298 MB | NPU
69
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS8750 | 5.541 ms | 1 - 292 MB | NPU
70
- | RTMDet | ONNX | w8a16 | Qualcomm® QCS7181 | 12.043 ms | 147 - 147 MB | NPU
71
- | RTMDet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 9.729 ms | 0 - 273 MB | NPU
72
- | RTMDet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 34.412 ms | 0 - 335 MB | NPU
73
- | RTMDet | TFLITE | float | Qualcomm® QCS8275 | 79.414 ms | 0 - 201 MB | NPU
74
- | RTMDet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 12.745 ms | 0 - 3 MB | NPU
75
- | RTMDet | TFLITE | float | Qualcomm® SA8775P | 114.326 ms | 3 - 33 MB | GPU
76
- | RTMDet | TFLITE | float | Qualcomm® SA8650P | 114.326 ms | 3 - 33 MB | GPU
77
- | RTMDet | TFLITE | float | Qualcomm® SA8255P | 114.326 ms | 3 - 33 MB | GPU
78
- | RTMDet | TFLITE | float | Qualcomm® QCS8450 | 34.412 ms | 0 - 335 MB | NPU
79
- | RTMDet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.352 ms | 0 - 201 MB | NPU
80
- | RTMDet | TFLITE | float | Qualcomm® SA7255P | 79.414 ms | 0 - 201 MB | NPU
81
- | RTMDet | TFLITE | float | Qualcomm® QCS9075 | 21.529 ms | 0 - 61 MB | NPU
82
- | RTMDet | TFLITE | float | Snapdragon® 8 Elite Mobile | 7.476 ms | 0 - 203 MB | NPU
83
- | RTMDet | TFLITE | float | Qualcomm® SA8295P | 27.266 ms | 0 - 261 MB | NPU
84
- | RTMDet | TFLITE | float | Qualcomm® QCS8750 | 7.476 ms | 0 - 203 MB | NPU
85
 
86
  ## License
87
  * The license for the original implementation of RTMDet can be found
 
15
  RTMDet is a highly efficient model for real-time object detection,capable of predicting both the bounding boxes and classes of objects within an image.It is highly optimized for real-time applications, making it reliable for industrial and commercial use
16
 
17
  This is based on the implementation of RTMDet found [here](https://github.com/open-mmlab/mmdetection/tree/3.x/configs/rtmdet).
18
+ 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/v0.58.0/src/qai_hub_models/models/rtmdet) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  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.
21
 
22
  ## Getting Started
23
  Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
24
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.58.0/src/qai_hub_models/models/rtmdet) Python library to compile and export the model with your own:
25
  - Custom weights (e.g., fine-tuned checkpoints)
26
  - Custom input shapes
27
  - Target device and runtime configurations
28
 
29
+ See our repository for [RTMDet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.58.0/src/qai_hub_models/models/rtmdet) for usage instructions.
30
 
31
 
32
  ## Model Details
 
42
  ## Performance Summary
43
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
44
  |---|---|---|---|---|---|---
45
+ | RTMDet | ONNX | float | Snapdragon® X2 Elite | 8.112 ms | 5 - 5 MB | NPU
46
+ | RTMDet | ONNX | float | Snapdragon® X Elite | 15.376 ms | 51 - 51 MB | NPU
47
+ | RTMDet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 11.142 ms | 1 - 230 MB | NPU
48
+ | RTMDet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 45.678 ms | 5 - 299 MB | NPU
49
+ | RTMDet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 14.703 ms | 0 - 53 MB | NPU
50
+ | RTMDet | ONNX | float | Qualcomm® QCS8450 | 45.678 ms | 5 - 299 MB | NPU
51
+ | RTMDet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 23.323 ms | 5 - 12 MB | NPU
52
+ | RTMDet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.249 ms | 4 - 194 MB | NPU
53
+ | RTMDet | ONNX | float | Snapdragon® 8 Elite Mobile | 9.317 ms | 2 - 186 MB | NPU
54
+ | RTMDet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 9.317 ms | 2 - 186 MB | NPU
55
+ | RTMDet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 15.376 ms | 51 - 51 MB | NPU
56
+ | RTMDet | ONNX | w8a16 | Snapdragon® X2 Elite | 5.763 ms | 2 - 2 MB | NPU
57
+ | RTMDet | ONNX | w8a16 | Snapdragon® X Elite | 14.194 ms | 27 - 27 MB | NPU
58
+ | RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 8.872 ms | 3 - 358 MB | NPU
59
+ | RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 19.028 ms | 3 - 357 MB | NPU
60
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 64.432 ms | 2 - 5 MB | NPU
61
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 13.505 ms | 0 - 33 MB | NPU
62
+ | RTMDet | ONNX | w8a16 | Qualcomm® QCS8450 | 19.028 ms | 3 - 357 MB | NPU
63
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 16.965 ms | 2 - 5 MB | NPU
64
+ | RTMDet | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 22.37 ms | 3 - 336 MB | NPU
65
+ | RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 5.108 ms | 2 - 325 MB | NPU
66
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 148.891 ms | 5 - 325 MB | NPU
67
+ | RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 6.171 ms | 0 - 296 MB | NPU
68
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 22.37 ms | 3 - 336 MB | NPU
69
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 6.171 ms | 0 - 296 MB | NPU
70
+ | RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 14.194 ms | 27 - 27 MB | NPU
71
+ | RTMDet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 19.853 ms | 0 - 305 MB | NPU
72
+ | RTMDet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 50.892 ms | 0 - 365 MB | NPU
73
+ | RTMDet | TFLITE | float | Qualcomm® QCS8275 | 280.108 ms | 3 - 33 MB | GPU
74
+ | RTMDet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 28.883 ms | 0 - 2 MB | NPU
75
+ | RTMDet | TFLITE | float | Qualcomm® SA8775P | 36.478 ms | 0 - 226 MB | NPU
76
+ | RTMDet | TFLITE | float | Qualcomm® SA8650P | 36.478 ms | 0 - 226 MB | NPU
77
+ | RTMDet | TFLITE | float | Qualcomm® SA8255P | 36.478 ms | 0 - 226 MB | NPU
78
+ | RTMDet | TFLITE | float | Qualcomm® QCS8450 | 50.892 ms | 0 - 365 MB | NPU
79
+ | RTMDet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 36.879 ms | 0 - 62 MB | NPU
80
+ | RTMDet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 9.885 ms | 0 - 230 MB | NPU
81
+ | RTMDet | TFLITE | float | Snapdragon® 8 Elite Mobile | 14.99 ms | 0 - 228 MB | NPU
82
+ | RTMDet | TFLITE | float | Qualcomm® SA7255P | 280.108 ms | 3 - 33 MB | GPU
83
+ | RTMDet | TFLITE | float | Qualcomm® SA8295P | 46.068 ms | 0 - 283 MB | NPU
84
+ | RTMDet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 14.99 ms | 0 - 228 MB | NPU
85
 
86
  ## License
87
  * The license for the original implementation of RTMDet can be found