Instructions to use Biodiversitas/sam3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Biodiversitas/sam3 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("mask-generation", model="Biodiversitas/sam3")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("Biodiversitas/sam3") model = AutoModel.from_pretrained("Biodiversitas/sam3", device_map="auto") - Notebooks
- Google Colab
- Kaggle
SAM 3 (mirror)
An unmodified copy of the SAM 3 weights by Meta, in the Hugging Face
Transformers format (Sam3Model, Sam3Processor). Every file is byte-identical to facebook/sam3
(e.g. model.safetensors SHA-256 6d06f0a5f84e…).
It is redistributed under the SAM License (see LICENSE), which permits redistribution provided a copy of the licence accompanies the materials. Use of these weights is subject to that licence, including its restrictions on military, weapons, nuclear and espionage uses and on trade-controlled parties.
This mirror is used by the Find subjects feature of Kunci Biodiversitas, which outlines every individual matching a text prompt (e.g. "crab", "orchid flower") in a photo dataset, so identification models can be trained on the subjects only. Inference runs locally on the user's computer.
Not affiliated with or endorsed by Meta. For the original model card, paper and code see facebook/sam3 and facebookresearch/sam3.
If you publish results obtained with SAM 3, acknowledge its use as the SAM License requires and cite:
Carion, N. et al. (2026). SAM 3: Segment Anything with Concepts. ICLR 2026. arXiv:2511.16719.
Maintained by PT Solusi Biodiversitas Nusantara (Sorong, Indonesia) · https://biodiversitas.id
- Downloads last month
- 19
Model tree for Biodiversitas/sam3
Base model
facebook/sam3