Update README.md
Browse files
README.md
CHANGED
|
@@ -1,149 +1,151 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
tags:
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
library_name: onnxruntime
|
| 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 |
-
urllib.request
|
| 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 |
-
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- image-segmentation
|
| 5 |
+
- segment-anything
|
| 6 |
+
- segment-anything-3
|
| 7 |
+
- open-vocabulary
|
| 8 |
+
- text-to-segmentation
|
| 9 |
+
- onnx
|
| 10 |
+
- onnxruntime
|
| 11 |
+
library_name: onnxruntime
|
| 12 |
+
base_model:
|
| 13 |
+
- facebook/sam3
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# Segment Anything 3 (SAM 3) — ONNX Models
|
| 17 |
+
|
| 18 |
+
ONNX-exported version of Meta's **Segment Anything Model 3 (SAM 3)**, an open-vocabulary segmentation model that accepts **text prompts** in addition to points and rectangles.
|
| 19 |
+
|
| 20 |
+
SAM 3 uses a CLIP-based language encoder to let you describe objects in natural language (e.g., `"truck"`, `"person with hat"`) and segment them without task-specific training.
|
| 21 |
+
|
| 22 |
+
These models are used by **[AnyLabeling](https://github.com/vietanhdev/anylabeling)** for AI-assisted image annotation, and exported by **[samexporter](https://github.com/vietanhdev/samexporter)**.
|
| 23 |
+
|
| 24 |
+
## Available Models
|
| 25 |
+
|
| 26 |
+
| File | Contents | Description |
|
| 27 |
+
|------|----------|-------------|
|
| 28 |
+
| `sam3_vit_h.zip` | 3 ONNX files | SAM 3 ViT-H (all components) |
|
| 29 |
+
|
| 30 |
+
The zip contains three ONNX components that work together:
|
| 31 |
+
|
| 32 |
+
| ONNX File | Role | Runs |
|
| 33 |
+
|-----------|------|------|
|
| 34 |
+
| `sam3_image_encoder.onnx` | Extracts visual features from the input image | Once per image |
|
| 35 |
+
| `sam3_language_encoder.onnx` | Encodes text prompt tokens into feature vectors | Once per text query |
|
| 36 |
+
| `sam3_decoder.onnx` | Produces segmentation masks given image + language features | Per prompt |
|
| 37 |
+
|
| 38 |
+
## Prompt Types
|
| 39 |
+
|
| 40 |
+
SAM 3 supports **three prompt modalities**:
|
| 41 |
+
|
| 42 |
+
| Prompt | Description |
|
| 43 |
+
|--------|-------------|
|
| 44 |
+
| **Text** | Natural-language description, e.g. `"truck"` — unique to SAM 3 |
|
| 45 |
+
| **Point** | Click `+point` / `-point` to include/exclude regions |
|
| 46 |
+
| **Rectangle** | Draw a bounding box around the target object |
|
| 47 |
+
|
| 48 |
+
Text prompts are the recommended workflow: they drive detection open-vocabulary style, so you can label **any object class** without retraining.
|
| 49 |
+
|
| 50 |
+
## Use with AnyLabeling (Recommended)
|
| 51 |
+
|
| 52 |
+
[AnyLabeling](https://github.com/vietanhdev/anylabeling) is a desktop annotation tool with a built-in model manager that downloads, caches, and runs these models automatically — no coding required.
|
| 53 |
+
|
| 54 |
+
1. Install: `pip install anylabeling`
|
| 55 |
+
2. Launch: `anylabeling`
|
| 56 |
+
3. Click the **Brain** button → select **Segment Anything 3 (ViT-H)** from the dropdown
|
| 57 |
+
4. Type a text description (e.g., `truck`) in the text prompt field
|
| 58 |
+
5. Optionally refine with point/rectangle prompts
|
| 59 |
+
|
| 60 |
+
[](https://github.com/vietanhdev/anylabeling)
|
| 61 |
+
|
| 62 |
+
## Use Programmatically with ONNX Runtime
|
| 63 |
+
|
| 64 |
+
```python
|
| 65 |
+
import urllib.request, zipfile
|
| 66 |
+
url = "https://huggingface.co/vietanhdev/segment-anything-3-onnx-models/resolve/main/sam3_vit_h.zip"
|
| 67 |
+
urllib.request.urlretrieve(url, "sam3_vit_h.zip")
|
| 68 |
+
with zipfile.ZipFile("sam3_vit_h.zip") as z:
|
| 69 |
+
z.extractall("sam3")
|
| 70 |
+
```
|
| 71 |
+
|
| 72 |
+
Then use [samexporter](https://github.com/vietanhdev/samexporter)'s inference module:
|
| 73 |
+
|
| 74 |
+
```bash
|
| 75 |
+
pip install samexporter
|
| 76 |
+
|
| 77 |
+
# Text prompt
|
| 78 |
+
python -m samexporter.inference \
|
| 79 |
+
--sam_variant sam3 \
|
| 80 |
+
--encoder_model sam3/sam3_image_encoder.onnx \
|
| 81 |
+
--decoder_model sam3/sam3_decoder.onnx \
|
| 82 |
+
--language_encoder_model sam3/sam3_language_encoder.onnx \
|
| 83 |
+
--image photo.jpg \
|
| 84 |
+
--prompt prompt.json \
|
| 85 |
+
--text_prompt "truck" \
|
| 86 |
+
--output result.png
|
| 87 |
+
```
|
| 88 |
+
|
| 89 |
+
Example `prompt.json` for a text-only query:
|
| 90 |
+
```json
|
| 91 |
+
[{"type": "text", "data": "truck"}]
|
| 92 |
+
```
|
| 93 |
+
|
| 94 |
+
## Model Architecture
|
| 95 |
+
|
| 96 |
+
SAM 3 follows the same encoder/decoder pattern as SAM and SAM 2, with an added CLIP-based language branch:
|
| 97 |
+
|
| 98 |
+
```
|
| 99 |
+
Input image ──► Image Encoder ──────────────────────────┐
|
| 100 |
+
▼
|
| 101 |
+
Text prompt ──► Language Encoder ──► Decoder ──► Masks + Scores + Boxes
|
| 102 |
+
▲
|
| 103 |
+
Optional: point / box prompts ──────────┘
|
| 104 |
+
```
|
| 105 |
+
|
| 106 |
+
The **image encoder** runs once per image and caches features. The **language encoder** runs once per text query. The **decoder** is lightweight and runs interactively for each prompt combination.
|
| 107 |
+
|
| 108 |
+
## Re-export from Source
|
| 109 |
+
|
| 110 |
+
To re-export or customize the models using [samexporter](https://github.com/vietanhdev/samexporter):
|
| 111 |
+
|
| 112 |
+
```bash
|
| 113 |
+
pip install samexporter
|
| 114 |
+
|
| 115 |
+
# Export all three SAM 3 ONNX components
|
| 116 |
+
python -m samexporter.export_sam3 --output_dir output_models/sam3
|
| 117 |
+
|
| 118 |
+
# Or use the convenience script:
|
| 119 |
+
bash convert_sam3.sh
|
| 120 |
+
```
|
| 121 |
+
|
| 122 |
+
## Custom Model Config for AnyLabeling
|
| 123 |
+
|
| 124 |
+
To use a locally re-exported SAM 3 as a custom model in AnyLabeling, create a `config.yaml`:
|
| 125 |
+
|
| 126 |
+
```yaml
|
| 127 |
+
type: segment_anything
|
| 128 |
+
name: sam3_vit_h_custom
|
| 129 |
+
display_name: Segment Anything 3 (ViT-H)
|
| 130 |
+
encoder_model_path: sam3_image_encoder.onnx
|
| 131 |
+
decoder_model_path: sam3_decoder.onnx
|
| 132 |
+
language_encoder_path: sam3_language_encoder.onnx
|
| 133 |
+
input_size: 1008
|
| 134 |
+
max_height: 1008
|
| 135 |
+
max_width: 1008
|
| 136 |
+
```
|
| 137 |
+
|
| 138 |
+
Then load it via **Brain button → Load Custom Model** in AnyLabeling.
|
| 139 |
+
|
| 140 |
+
## Related Repositories
|
| 141 |
+
|
| 142 |
+
| Repo | Description |
|
| 143 |
+
|------|-------------|
|
| 144 |
+
| [vietanhdev/samexporter](https://github.com/vietanhdev/samexporter) | Export scripts, inference code, conversion tools |
|
| 145 |
+
| [vietanhdev/anylabeling](https://github.com/vietanhdev/anylabeling) | Desktop annotation app powered by these models |
|
| 146 |
+
| [facebook/sam3](https://huggingface.co/facebook/sam3) | Original SAM 3 PyTorch checkpoint by Meta |
|
| 147 |
+
|
| 148 |
+
## License
|
| 149 |
+
|
| 150 |
+
The ONNX models are derived from Meta's SAM 3, released under the **[SAM License](https://github.com/facebookresearch/sam3/blob/main/LICENSE)**.
|
| 151 |
+
The export code is part of [samexporter](https://github.com/vietanhdev/samexporter), released under the **MIT** license.
|