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
78cf0a0
·
verified ·
1 Parent(s): 2efa791

v0.57.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.57.0 for changelog.

Files changed (2) hide show
  1. LICENSE +1 -0
  2. README.md +96 -0
LICENSE ADDED
@@ -0,0 +1 @@
 
 
1
+ The license of the original trained model can be found at https://github.com/facebookresearch/sam3/blob/main/LICENSE.
README.md ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ license: other
4
+ tags:
5
+ - foundation
6
+ - android
7
+ pipeline_tag: image-segmentation
8
+
9
+ ---
10
+
11
+ ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sam3/web-assets/model_demo.png)
12
+
13
+ # Segment-Anything-Model-3: Optimized for Qualcomm Devices
14
+
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/v0.57.0/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/v0.57.0/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/v0.57.0/src/qai_hub_models/models/sam3) for usage instructions.
30
+
31
+
32
+ ## Model Details
33
+
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 | 211.492 ms | 135 - 135 MB | NPU
48
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 484.733 ms | 667 - 667 MB | NPU
49
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 381.125 ms | 156 - 167 MB | NPU
50
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 236.184 ms | 131 - 143 MB | NPU
51
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS9075 | 503.413 ms | 106 - 215 MB | NPU
52
+ | head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 286.171 ms | 132 - 143 MB | NPU
53
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8750 | 286.171 ms | 132 - 143 MB | NPU
54
+ | head | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS7181 | 484.733 ms | 667 - 667 MB | NPU
55
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 210.844 ms | 106 - 106 MB | NPU
56
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 489.544 ms | 107 - 107 MB | NPU
57
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 385.297 ms | 106 - 118 MB | NPU
58
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8275 | 1006.597 ms | 87 - 95 MB | NPU
59
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 236.135 ms | 81 - 90 MB | NPU
60
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 1006.597 ms | 87 - 95 MB | NPU
61
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® QCS9075 | 503.79 ms | 106 - 278 MB | NPU
62
+ | head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 279.471 ms | 81 - 94 MB | NPU
63
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 596.634 ms | 87 - 96 MB | NPU
64
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8750 | 279.471 ms | 81 - 94 MB | NPU
65
+ | head | QNN_CONTEXT_BINARY | float | Qualcomm® QCS7181 | 489.544 ms | 107 - 107 MB | NPU
66
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1077.436 ms | 265 - 265 MB | NPU
67
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 2157.567 ms | 944 - 944 MB | NPU
68
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1721.251 ms | 20 - 31 MB | NPU
69
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1208.259 ms | 56 - 68 MB | NPU
70
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS9075 | 2431.819 ms | 117 - 131 MB | NPU
71
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 1405.929 ms | 18 - 29 MB | NPU
72
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8750 | 1405.929 ms | 18 - 29 MB | NPU
73
+ | vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS7181 | 2157.567 ms | 944 - 944 MB | NPU
74
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 1078.092 ms | 12 - 12 MB | NPU
75
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 2190.203 ms | 12 - 12 MB | NPU
76
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 1731.302 ms | 12 - 27 MB | NPU
77
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8275 | 6355.089 ms | 8 - 17 MB | NPU
78
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 1230.139 ms | 24 - 33 MB | NPU
79
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 6355.089 ms | 8 - 17 MB | NPU
80
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® QCS9075 | 2430.561 ms | 12 - 131 MB | NPU
81
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 1397.335 ms | 0 - 9 MB | NPU
82
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 2813.386 ms | 0 - 9 MB | NPU
83
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8750 | 1397.335 ms | 0 - 9 MB | NPU
84
+ | vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® QCS7181 | 2190.203 ms | 12 - 12 MB | NPU
85
+
86
+ ## License
87
+ * The license for the original implementation of Segment-Anything-Model-3 can be found
88
+ [here](https://github.com/facebookresearch/sam3/blob/main/LICENSE).
89
+
90
+ ## References
91
+ * [SAM 3: Segment Anything with Concepts](https://arxiv.org/abs/2511.16719)
92
+ * [Source Model Implementation](https://github.com/facebookresearch/sam3)
93
+
94
+ ## Community
95
+ * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
96
+ * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).