SAM3-LiteText: An Anatomical Study of the SAM3 Text Encoder for Efficient Vision-Language Segmentation
Paper
• 2602.12173 • Published
• 1
This is a Transformers-compatible checkpoint of the SAM3-LiteText paper.
Usage is identical to SAM-3 in Transformers: https://huggingface.co/docs/transformers/main/en/model_doc/sam3#usage-examples-with--transformers.
You just need to replace the model by:
# make sure to load from this branch as long as it isn't merged: https://github.com/nielsrogge/transformers/add_sam_3_lite_text
from transformers import Sam3LiteTextModel
model = Sam3LiteTextModel.from_pretrained("nielsr/efficient-sam-3-mobileclip-s0-ctx16")
PR is here: https://github.com/huggingface/transformers/pull/44320.