Hugging Face's logo

qualcomm
/
Segment-Anything-Model-3

Image Segmentation
PyTorch
foundation
android
Model card Files
xet
 Community
1
qaihm-bot commited on
Commit
8135458
·
verified ·
1 Parent(s): 6d569c7

v0.60.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.60.0 for changelog.

Files changed (1) hide show
  1. README.md +52 -50
README.md CHANGED
@@ -15,18 +15,18 @@ pipeline_tag: image-segmentation
15
  SAM3 (Segment Anything with Concepts) extends SAM2 with open-vocabulary segmentation, producing bounding boxes and masks for objects matching a natural-language prompt.
16
 
17
  This is based on the implementation of Segment-Anything-Model-3 found [here](https://github.com/facebookresearch/sam3).
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/v0.59.0/src/qai_hub_models/models/sam3) 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/v0.59.0/src/qai_hub_models/models/sam3) 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 [Segment-Anything-Model-3 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/sam3) for usage instructions.
30
 
31
 
32
  ## Model Details
@@ -34,60 +34,62 @@ See our repository for [Segment-Anything-Model-3 on GitHub](https://github.com/q
34
  **Model Type:** Model_use_case.semantic_segmentation
35
 
36
  **Model Stats:**
37
- - Model checkpoint: sam3
38
  - Input resolution: 1008x1008
39
- - Number of parameters (SAM3Backbone): 33.5M
40
  - Model size (SAM3Backbone) (float): 128 MB
41
- - Number of parameters (SAM3Transformer): 6.22M
42
  - Model size (SAM3Transformer) (float): 23.7 MB
 
 
43
 
44
  ## Performance Summary
45
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
46
  |---|---|---|---|---|---|---
47
- | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 209.595 ms | 136 - 136 MB | NPU
48
- | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 489.038 ms | 667 - 667 MB | NPU
49
- | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 389.841 ms | 155 - 168 MB | NPU
50
- | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 643.219 ms | 106 - 215 MB | NPU
51
- | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 489.038 ms | 667 - 667 MB | NPU
52
- | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 282.536 ms | 131 - 143 MB | NPU
53
- | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 282.536 ms | 131 - 143 MB | NPU
54
- | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 229.69 ms | 131 - 144 MB | NPU
55
- | head | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 211.571 ms | 107 - 107 MB | NPU
56
- | head | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 484.374 ms | 106 - 106 MB | NPU
57
- | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 384.956 ms | 105 - 117 MB | NPU
58
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8275 | 1006.314 ms | 106 - 116 MB | NPU
59
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 517.235 ms | 106 - 116 MB | NPU
60
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 517.235 ms | 106 - 116 MB | NPU
61
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 517.235 ms | 106 - 116 MB | NPU
62
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 502.762 ms | 106 - 278 MB | NPU
63
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 484.374 ms | 106 - 106 MB | NPU
64
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 270.105 ms | 81 - 94 MB | NPU
65
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 1006.314 ms | 106 - 116 MB | NPU
66
- | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 597.053 ms | 106 - 111 MB | NPU
67
- | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 270.105 ms | 81 - 94 MB | NPU
68
- | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 242.095 ms | 35 - 44 MB | NPU
69
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1095.36 ms | 265 - 265 MB | NPU
70
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 2202.976 ms | 943 - 943 MB | NPU
71
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1709.651 ms | 20 - 31 MB | NPU
72
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 2433.358 ms | 12 - 26 MB | NPU
73
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2202.976 ms | 943 - 943 MB | NPU
74
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1427.598 ms | 8 - 20 MB | NPU
75
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 1427.598 ms | 8 - 20 MB | NPU
76
- | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1234.583 ms | 28 - 40 MB | NPU
77
- | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 1076.267 ms | 12 - 12 MB | NPU
78
- | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 2180.381 ms | 12 - 12 MB | NPU
79
- | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 1687.452 ms | 12 - 24 MB | NPU
80
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8275 | 6386.65 ms | 12 - 21 MB | NPU
81
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 2606.394 ms | 1 - 10 MB | NPU
82
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 2606.394 ms | 1 - 10 MB | NPU
83
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 2606.394 ms | 1 - 10 MB | NPU
84
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 2430.088 ms | 12 - 131 MB | NPU
85
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 2180.381 ms | 12 - 12 MB | NPU
86
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 1412.697 ms | 0 - 13 MB | NPU
87
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 6386.65 ms | 12 - 21 MB | NPU
88
- | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 2814.768 ms | 1 - 11 MB | NPU
89
- | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 1412.697 ms | 0 - 13 MB | NPU
90
- | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 1211.98 ms | 5 - 15 MB | NPU
 
 
91
 
92
  ## License
93
  * The license for the original implementation of Segment-Anything-Model-3 can be found
 
15
  SAM3 (Segment Anything with Concepts) extends SAM2 with open-vocabulary segmentation, producing bounding boxes and masks for objects matching a natural-language prompt.
16
 
17
  This is based on the implementation of Segment-Anything-Model-3 found [here](https://github.com/facebookresearch/sam3).
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/v0.60.0/src/qai_hub_models/models/sam3) 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/v0.60.0/src/qai_hub_models/models/sam3) 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 [Segment-Anything-Model-3 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/sam3) for usage instructions.
30
 
31
 
32
  ## Model Details
 
34
  **Model Type:** Model_use_case.semantic_segmentation
35
 
36
  **Model Stats:**
 
37
  - Input resolution: 1008x1008
38
+ - Model checkpoint: sam3
39
  - Model size (SAM3Backbone) (float): 128 MB
 
40
  - Model size (SAM3Transformer) (float): 23.7 MB
41
+ - Number of parameters (SAM3Backbone): 33.5M
42
+ - Number of parameters (SAM3Transformer): 6.22M
43
 
44
  ## Performance Summary
45
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
46
  |---|---|---|---|---|---|---
47
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 220.565 ms | 136 - 136 MB | NPU
48
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 484.226 ms | 667 - 667 MB | NPU
49
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 384.365 ms | 156 - 168 MB | NPU
50
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 492.325 ms | 106 - 216 MB | NPU
51
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 503.652 ms | 106 - 215 MB | NPU
52
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 484.226 ms | 667 - 667 MB | NPU
53
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 281.628 ms | 132 - 144 MB | NPU
54
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 281.628 ms | 132 - 144 MB | NPU
55
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 240.153 ms | 131 - 143 MB | NPU
56
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 210.868 ms | 107 - 107 MB | NPU
57
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 484.956 ms | 106 - 106 MB | NPU
58
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 383.366 ms | 85 - 92 MB | NPU
59
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 1006.102 ms | 100 - 109 MB | NPU
60
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 518.153 ms | 106 - 117 MB | NPU
61
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 518.153 ms | 106 - 117 MB | NPU
62
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 518.153 ms | 106 - 117 MB | NPU
63
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 503.463 ms | 106 - 278 MB | NPU
64
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 484.956 ms | 106 - 106 MB | NPU
65
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 279.29 ms | 81 - 95 MB | NPU
66
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 1006.102 ms | 100 - 109 MB | NPU
67
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 597.358 ms | 106 - 112 MB | NPU
68
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 279.29 ms | 81 - 95 MB | NPU
69
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 232.673 ms | 30 - 38 MB | NPU
70
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1082.695 ms | 265 - 265 MB | NPU
71
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 2227.915 ms | 943 - 943 MB | NPU
72
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1736.339 ms | 128 - 138 MB | NPU
73
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 2530.033 ms | 50 - 65 MB | NPU
74
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 2455.828 ms | 117 - 131 MB | NPU
75
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2227.915 ms | 943 - 943 MB | NPU
76
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1444.345 ms | 10 - 21 MB | NPU
77
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 1444.345 ms | 10 - 21 MB | NPU
78
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1238.306 ms | 8 - 20 MB | NPU
79
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 1080.835 ms | 12 - 12 MB | NPU
80
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 2189.109 ms | 12 - 12 MB | NPU
81
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 1735.446 ms | 12 - 18 MB | NPU
82
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 6377.315 ms | 2 - 12 MB | NPU
83
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 2520.691 ms | 1 - 12 MB | NPU
84
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 2520.691 ms | 1 - 12 MB | NPU
85
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 2520.691 ms | 1 - 12 MB | NPU
86
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 2814.293 ms | 12 - 131 MB | NPU
87
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 2189.109 ms | 12 - 12 MB | NPU
88
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 1417.108 ms | 0 - 13 MB | NPU
89
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 6377.315 ms | 2 - 12 MB | NPU
90
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 2784.976 ms | 9 - 19 MB | NPU
91
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 1417.108 ms | 0 - 13 MB | NPU
92
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 1203.763 ms | 3 - 15 MB | NPU
93
 
94
  ## License
95
  * The license for the original implementation of Segment-Anything-Model-3 can be found