Download Checkpoints and Assets

All commands below assume you are in the WorldAtlas checkout and that WORLDATLAS_CHECKPOINT_ROOT points at the active checkpoint root when you use a shared disk layout.

One-click download

bash scripts/download_metric_checkpoints.sh          # core metric + official runtime weights
bash scripts/download_metric_checkpoints.sh --all    # + FVMD, JEDi, TRAJAN, HPSv3, TransNetV2, caches

Missing checkpoints raise FileNotFoundError during evaluation — WorldAtlas does not download weights online at score time.

Metric ↔ checkpoint quick reference

Metric(s)	Path under ckpt/	Download source
Long sequence metrics, long-horizon diagnostics	AMT model amt-s.pth, RAFT model raft-things.pth, MUSIQ SPAQ checkpoint	HF: lalala125/AMT, RAFT Dropbox models.zip, IQA-PyTorch release
Temporal calibration, physical FPS	Pulse of Motion checkpoint	HF: xiangbog/Visual Chronometer
Camera error, reprojection error (DROID)	Official DROID weights	Google Drive via gdown
Motion accuracy	GroundingDINO, SAM, SAM 2.1, BERT base	GroundingDINO, SAM ViT-H, HF: facebook/sam2.1-hiera-base-plus, HF: google-bert/bert-base-uncased
Optical flow AEPE, motion magnitude, motion accuracy	SEA-RAFT weights, RAFT Things	HF: Yarimasune/SEA-RAFT, copy from the RAFT model directory
Motion smoothness	VFIMamba weights	HF: MCG-NJU/VFIMamba (rename the downloaded model file to VFIMamba.pkl)
Long sequence aesthetic quality	Aesthetic predictor weights	improved-aesthetic-predictor
Reconstruction consistency, met3r consistency	DA3-GIANT, MASt3R, DINO	HF: depth-anything/DA3-GIANT, HF: naver/MASt3R
Background consistency, memory revisit consistency	CLIP, DINOv2	CLIP, DINOv2
HPSv3	HPSv3 weights, config, Qwen2-VL	HF: MizzenAI/HPSv3, config template, HF: Qwen/Qwen2-VL-7B-Instruct
Segment continuity	TransNetV2 weights	soCzech/TransNetV2
FVMD (video suite)	PIPs++ weights	DSL-Lab FVMD PIPs++ weights
JEDi (video suite)	JEDi ViT-H and SSv2 probe	Meta JEPA weights
TRAJAN (dynamic suites)	BootsTAPIR checkpoint, track autoencoder	Google DeepMind BootsTAPIR, TRAJAN autoencoder

The in-tree adapters resolve files from the official metrics directory at evaluation time. Exact filenames are in the download commands below.

1. Benchmark metric weights

CKPT="${WORLDATLAS_CHECKPOINT_ROOT:-ckpt}"

mkdir -p "$CKPT"/{amt_model,raft_model/models,Pulse-of-Motion,musiq,official_metrics/models}

wget -q -O "$CKPT/amt_model/amt-s.pth" \
  https://huggingface.co/lalala125/AMT/resolve/main/amt-s.pth

wget -q -P /tmp https://dl.dropboxusercontent.com/s/4j4z58wuv8o0mfz/models.zip
unzip -o /tmp/models.zip -d "$CKPT/raft_model"

python -m pip install -q huggingface_hub
python - <<'PY'
import os
from huggingface_hub import hf_hub_download
root = os.environ.get("WORLDATLAS_CHECKPOINT_ROOT", "ckpt")
path = hf_hub_download(
    repo_id="xiangbog/Visual_Chronometer",
    filename="vc_common_10_60fps.ckpt",
    local_dir=f"{root}/Pulse-of-Motion",
)
print(path)
PY

wget -q -O "$CKPT/musiq/musiq_spaq_ckpt-358bb6af.pth" \
  https://github.com/chaofengc/IQA-PyTorch/releases/download/v0.1-weights/musiq_spaq_ckpt-358bb6af.pth

2. Official metric runtime weights

These weights back motion_accuracy and the official camera/flow metrics:

CKPT="${WORLDATLAS_CHECKPOINT_ROOT:-ckpt}"
export CKPT

wget -q -P "$CKPT/official_metrics" \
  https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth

wget -q -P "$CKPT/official_metrics" \
  https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

wget -q -P "$CKPT/official_metrics" \
  https://huggingface.co/facebook/sam2.1-hiera-base-plus/resolve/main/sam2.1_hiera_base_plus.pt

cp "$CKPT/raft_model/models/raft-things.pth" "$CKPT/official_metrics/models/raft-things.pth"

python -m pip install -q gdown
gdown 1PpqVt1H4maBa_GbPJp4NwxRsd9jk-elh -O "$CKPT/official_metrics/droid.pth"

python - <<'PY'
import os
import shutil
from huggingface_hub import hf_hub_download

root = os.environ["CKPT"]
official = f"{root}/official_metrics"

tartan_src = hf_hub_download(
    repo_id="Yarimasune/SEA-RAFT",
    filename="Tartan-C-T-TSKH-spring540x960-M.pth",
)
shutil.copy2(tartan_src, f"{official}/Tartan-C-T-TSKH-spring540x960-M.pth")

vfimamba_src = hf_hub_download(repo_id="MCG-NJU/VFIMamba", filename="model.pkl")
shutil.copy2(vfimamba_src, f"{official}/VFIMamba.pkl")
print(tartan_src)
print(vfimamba_src)
PY

wget -q -O "$CKPT/official_metrics/sac+logos+ava1-l14-linearMSE.pth" \
  https://github.com/christophschuhmann/improved-aesthetic-predictor/raw/main/sac%2Blogos%2Bava1-l14-linearMSE.pth

if command -v huggingface-cli >/dev/null 2>&1; then
  export HF_HOME="$CKPT/huggingface"
  export HUGGINGFACE_HUB_CACHE="$CKPT/huggingface/hub"
  huggingface-cli download google-bert/bert-base-uncased --local-dir "$CKPT/official_metrics/bert-base-uncased"
else
  python - <<'PY'
from huggingface_hub import snapshot_download
import os
snapshot_download(
    repo_id="google-bert/bert-base-uncased",
    local_dir=os.path.join(os.environ["CKPT"], "official_metrics", "bert-base-uncased"),
)
PY
fi

3. Quality, distribution, and trajectory metric extras

Required when running hpsv3_norm, segment_continuity, or the video-suite jedi metric:

CKPT="${WORLDATLAS_CHECKPOINT_ROOT:-ckpt}"

# HPSv3 reward model + Qwen2-VL backbone
wget -q -O "$CKPT/HPSv3/HPSv3.safetensors" \
  https://huggingface.co/MizzenAI/HPSv3/resolve/main/HPSv3.safetensors
wget -q -O "$CKPT/HPSv3/HPSv3_7B_local.yaml" \
  https://raw.githubusercontent.com/MizzenAI/HPSv3/main/hpsv3/config/HPSv3_7B.yaml
huggingface-cli download Qwen/Qwen2-VL-7B-Instruct --local-dir "$CKPT/Qwen2-VL-7B-Instruct"

# JEDi / V-JEPA (video distribution metric)
wget -q -O "$CKPT/JEDi/vith16.pth.tar" \
  https://dl.fbaipublicfiles.com/jepa/vith16/vith16.pth.tar
wget -q -O "$CKPT/JEDi/ssv2-probe.pth.tar" \
  https://dl.fbaipublicfiles.com/jepa/vith16/ssv2-probe.pth.tar

# FVMD / PIPs++ (video distribution metric)
wget -q -O "$CKPT/torch/hub/checkpoints/pips2_weights.pth" \
  https://github.com/DSL-Lab/FVMD-frechet-video-motion-distance/releases/download/pips2_weights/pips2_weights.pth

# TRAJAN / BootsTAPIR (per-video dynamic metric)
wget -q -O "$CKPT/TRAJAN/bootstapir_checkpoint_v2.npy" \
  https://storage.googleapis.com/dm-tapnet/bootstap/bootstapir_checkpoint_v2.npy
wget -q -O "$CKPT/TRAJAN/track_autoencoder_ckpt.npz" \
  https://storage.googleapis.com/dm-tapnet/trajan/track_autoencoder_ckpt.npz

# TransNetV2 (segment_continuity): clone upstream, pull LFS weights, convert to PyTorch
git clone --depth 1 https://github.com/soCzech/TransNetV2 thirdparty/TransNetV2
git -C thirdparty/TransNetV2 lfs pull
cd thirdparty/TransNetV2/inference-pytorch && python convert_weights.py
cp transnetv2-pytorch-weights.pth "$CKPT/TransNetV2/"

Or run the bundled helper:

bash scripts/download_metric_checkpoints.sh --all

4. Hugging Face and Torch caches

WorldAtlas pins HF/Torch caches under the checkpoint root and runs bundled adapters offline. Preload the backends you need before evaluation:

CKPT="${WORLDATLAS_CHECKPOINT_ROOT:-ckpt}"

export HF_HOME="$CKPT/huggingface"
export HUGGINGFACE_HUB_CACHE="$CKPT/huggingface/hub"
export TRANSFORMERS_CACHE="$CKPT/transformers"
export TORCH_HOME="$CKPT/torch"

huggingface-cli download openai/clip-vit-base-patch16
huggingface-cli download facebook/dinov2-base
huggingface-cli download depth-anything/DA3-GIANT --local-dir "$CKPT/DA3-GIANT"
huggingface-cli download naver/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric

python - <<'PY'
import os
import torch
torch.hub.set_dir(os.environ.get("TORCH_HOME", "ckpt/torch"))
torch.hub.load("facebookresearch/dino:main", "dino_deitsmall16")
PY

Replace the Python one-liner with a small script if you prefer explicit shell exports; the important constraint is that DINO weights land under ckpt/torch/hub/.