| --- |
| license: other |
| license_name: sam-license |
| license_link: https://github.com/facebookresearch/sam3/blob/main/LICENSE |
| language: |
| - en |
| base_model: facebook/sam3 |
| pipeline_tag: image-segmentation |
| tags: |
| - darktable |
| - onnx |
| - sam3 |
| - image-segmentation |
| --- |
| |
| # SAM 3 for darktable |
|
|
| This repository provides a prebuilt darktable `.dtmodel` package containing SAM3 weights converted to ONNX for interactive object masking. |
|
|
| ## Requirements |
|
|
| - darktable 5.8 or newer |
|
|
| ## Installation |
|
|
| 1. Download [`mask-object-sam3.dtmodel`](dtmodels/mask-object-sam3.dtmodel). |
| 2. Import it using darktable's local AI model installer. |
| 3. Select `mask sam3` for object masking. |
|
|
| ## Source |
|
|
| Original model: |
|
|
| - Segment Anything Model 3 by Meta AI / Facebook Research |
| - Repository: https://github.com/facebookresearch/sam3 |
| - Hugging Face: https://huggingface.co/facebook/sam3 |
| - Paper: https://arxiv.org/abs/2511.16719 |
|
|
| ## License |
|
|
| SAM3 and the converted ONNX weights are distributed under the custom [SAM License](https://github.com/facebookresearch/sam3/blob/main/LICENSE). |
|
|
| By using or redistributing the package, you agree to its terms. Read the complete license before use or redistribution. |
|
|
| This conversion and package are not endorsed by Meta, Facebook Research, Hugging Face, or darktable. Darktable does not distribute or endorse SAM3. |
|
|
| ## Guide |
|
|
| The steps are explained in this video: |
|
|
| [YouTube](https://youtu.be/Tubz-noyfM4) |
|
|
