Skip to content

README.md

Latest commit

 

History

History
25 lines (19 loc) · 1.33 KB

README.md

File metadata and controls

25 lines (19 loc) · 1.33 KB
Raw

Model checkpoints

The pretrained checkpoints are not committed to the repository (they are large and gitignored). Download them into this weights/ directory before running the pipeline.

File Model Size Source
CropFormer_hornet_3x_03823a.pth CropFormer (HorNet, 3x) — the 2D class-agnostic mask model used in the paper ~849 MB Adobe_EntitySeg on Hugging Face
sam_vit_h_4b8939.pth Segment Anything (ViT-H) — used for the OpenMask3D-style CLIP feature extraction ~2.4 GB segment-anything

Download

# SAM ViT-H
wget -P weights https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

# CropFormer HorNet 3x (from the Adobe_EntitySeg Hugging Face repo)
wget -O weights/CropFormer_hornet_3x_03823a.pth \
  "https://huggingface.co/datasets/qqlu1992/Adobe_EntitySeg/resolve/main/CropFormer_model/Entity_Segmentation/CropFormer_hornet_3x/CropFormer_hornet_3x_03823a.pth"

The default paths used by the code are:

  • weights/CropFormer_hornet_3x_03823a.pth (see ovmap/build_map.py:get_cf_setting)
  • weights/sam_vit_h_4b8939.pth (--sam_checkpoint_path)