Update README.md

README.md

@@ -22,53 +22,57 @@ A benchmark for evaluating 3D point motion in video, covering egocentric and thi
 
 Benchmark data provided in this repository include annotations, captions, indices, and scripts created by Ai2 that correspond to the source datasets.
 
-```bash
-# pip install huggingface_hub
-# If the download stalls near completion: HF_HUB_DISABLE_XET=1 huggingface-cli download ...
-huggingface-cli download allenai/PointMotionBench \
-    --repo-type dataset --local-dir $POINTMOTIONBENCH_ROOT
-```
-
----
-
-### Step 2 — DAVIS: Download Videos
 
-DAVIS videos should be reconstructed from the source data at the [official source](https://davischallenge.org/davis2017/code.html). Download **Trainval 2017 - Images (480p)** and **Annotations**, then convert frames to mp4 (requires `ffmpeg`):
-
-```bash
-python davis/reconstruct_davis.py \
-    --davis-root /path/to/DAVIS \
-    --output-dir davis/videos/input_480p
+```python
+# pip install huggingface_hub
+# If the download stalls near completion: HF_HUB_DISABLE_XET=1 python your_script.py
+from huggingface_hub import snapshot_download
+snapshot_download(
+    repo_id="allenai/PointMotionBench",
+    repo_type="dataset",
+    local_dir=".",
+)
 ```
 
 ---
 
-### Step 3 — HOT3D: Download Videos
+### Step 2 — HOT3D: Download Videos
 
-HOT3D videos should be reconstructed from the source data at [bop-benchmark/hot3d](https://huggingface.co/datasets/bop-benchmark/hot3d) (HuggingFace).
+We do not share videos from HOT3D. Users should access the videos from the source dataset at [bop-benchmark/hot3d](https://huggingface.co/datasets/bop-benchmark/hot3d) (HuggingFace).
 
-**Requirements:** `huggingface_hub`, `imageio[ffmpeg]`, `imageio-ffmpeg`, `opencv-python-headless`, `numpy`
+**Requirements:** `imageio[ffmpeg]`, `imageio-ffmpeg`, `opencv-python-headless`, `numpy`
 
 ```bash
-python hot3d/reconstruct_hot3d.py \
-    --workdir    /path/to/hot3d_work \
-    --output-dir hot3d/rgbs
+# 1. download train_aria TARs (~1,516 clips)
+#    to download only the 1,272 clips needed for PointMotionBench, add:
+#    --captions hot3d/hot3d_annotations.json
+python hot3d/download_train_aria.py --output /path/to/train_aria
+
+# 2. extract undistorted upright RGB videos (one mp4 per TAR)
+python hot3d/extract_rgbs.py \
+    --clips_dir  /path/to/train_aria \
+    --output_dir /path/to/rgbs
+
+# 3. trim to PointMotionBench windows
+python hot3d/trim_hot3d_clips.py \
+    --src_dir    /path/to/rgbs \
+    --captions   hot3d/hot3d_annotations.json \
+    --output_dir hot3d/videos
 ```
 
-This runs all three stages (download TARs → extract RGB → trim to PointMotionBench windows).
-For large-scale extraction, run the three scripts individually — `extract_rgbs.py` supports sharding:
+For large-scale extraction, `extract_rgbs.py` supports sharding:
 
 ```bash
 python hot3d/extract_rgbs.py \
-    --clips_dir  /path/to/hot3d_work/train_aria \
-    --output_dir /path/to/hot3d_work/rgbs \
+    --clips_dir  /path/to/train_aria \
+    --output_dir /path/to/rgbs \
     --shard_idx  0 \
     --num_shards 8
 ```
 
 ---
 
-### Step 4 — WorldTrack: Reconstruct Clips
+### Step 3 — WorldTrack: extract clips
 
 Download the WorldTrack source data (WorldTrack benchmark, introduced in St4RTrack, Feng et al., ICCV 2025 — dataset download available at [HavenFeng/St4RTrack](https://github.com/HavenFeng/St4RTrack)). The source data should have this layout:
 
@@ -83,7 +87,7 @@ WorldTrack/
 Then extract PointMotionBench clips using the index map from Step 1:
 
 ```bash
-python worldtrack/reconstruct_worldtrack.py \
+python worldtrack/extract_worldtrack_clips.py \
     --index_map  worldtrack/worldtrack_index_map.json \
     --src_dir    /path/to/WorldTrack \
     --output_dir worldtrack
@@ -107,16 +111,3 @@ Such videos and data are provided by the owners of the source datasets above, an
 license terms and use restrictions. Users who access videos and data from these sources are responsible for 
 reviewing and confirming that their use complies with the terms and conditions.
 
-## Citation
-
-```bibtex
-@misc{zhang2026molmomotionforecastingpointtrajectories,
-    title={MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction}, 
-    author={Jianing Zhang and Chenhao Zheng and Yajun Yang and Max Argus and Rustin Soraki and Winson Han and Taira Anderson and Chun-Liang Li and Shuo Liu and Jiafei Duan and Zhongzheng Ren and Jieyu Zhang and Ranjay Krishna},
-    year={2026},
-    eprint={2606.18558},
-    archivePrefix={arXiv},
-    primaryClass={cs.CV},
-    url={https://arxiv.org/abs/2606.18558}, 
-}
-```

davis/reconstruct_davis.py

@@ -1,141 +0,0 @@
-#!/usr/bin/env python3
-"""
-Reconstruct DAVIS PointMotionBench videos from DAVIS 2017 Trainval frames.
-
-Download Trainval 2017 - Images (480p) and Annotations from
-https://davischallenge.org/davis2017/code.html, then run this script.
-
-The DAVIS root should have this layout after extraction:
-    <davis-root>/
-    └── JPEGImages/
-        └── 480p/
-            └── <sequence>/
-                ├── 00000.jpg
-                ├── 00001.jpg
-                └── ...
-
-Example:
-    python davis/reconstruct_davis.py \
-        --davis-root /path/to/DAVIS \
-        --output-dir davis/videos/input_480p \
-        --captions   davis/davis_captions.json
-"""
-
-import argparse
-import json
-import os
-import shutil
-import subprocess
-import sys
-from pathlib import Path
-
-TARGET_FPS = 24
-
-
-def ffmpeg_exe():
-    if shutil.which("ffmpeg"):
-        return "ffmpeg"
-    try:
-        import imageio_ffmpeg
-        return imageio_ffmpeg.get_ffmpeg_exe()
-    except Exception:
-        raise RuntimeError(
-            "ffmpeg not found; install a system ffmpeg or `pip install imageio-ffmpeg`."
-        )
-
-
-def encode_sequence(ffmpeg, frames_dir, dst, fps):
-    frames = sorted(frames_dir.glob("*.jpg"))
-    if not frames:
-        raise RuntimeError(f"no .jpg frames found in {frames_dir}")
-
-    dst.parent.mkdir(parents=True, exist_ok=True)
-    tmp = dst.with_suffix(".tmp.mp4")
-    cmd = [
-        ffmpeg, "-y", "-loglevel", "error",
-        "-framerate", str(fps),
-        "-i", str(frames_dir / "%05d.jpg"),
-        "-vf", "crop=trunc(iw/2)*2:trunc(ih/2)*2",
-        "-c:v", "libx264", "-crf", "18", "-preset", "fast",
-        "-pix_fmt", "yuv420p", "-an",
-        str(tmp),
-    ]
-    res = subprocess.run(cmd, capture_output=True)
-    if res.returncode != 0:
-        if tmp.exists():
-            tmp.unlink()
-        raise RuntimeError(res.stderr.decode(errors="replace").strip() or "ffmpeg failed")
-    os.replace(tmp, dst)
-
-
-def main():
-    parser = argparse.ArgumentParser(
-        description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter
-    )
-    parser.add_argument(
-        "--davis-root", required=True, type=Path,
-        help="DAVIS 2017 root dir containing JPEGImages/480p/"
-    )
-    parser.add_argument(
-        "--output-dir", type=Path,
-        default=Path(__file__).resolve().parent / "videos" / "input_480p",
-        help="Output directory for mp4 files (default: davis/videos/input_480p/)",
-    )
-    parser.add_argument(
-        "--captions", type=Path,
-        default=Path(__file__).resolve().parent / "davis_captions.json",
-        help="davis_captions.json (default: alongside this script)",
-    )
-    parser.add_argument("--fps", type=int, default=TARGET_FPS)
-    parser.add_argument("--overwrite", action="store_true", help="Re-encode even if output exists")
-    args = parser.parse_args()
-
-    frames_root = args.davis_root / "JPEGImages" / "480p"
-    if not frames_root.exists():
-        print(f"ERROR: {frames_root} does not exist — check --davis-root points to the DAVIS 2017 root.")
-        sys.exit(1)
-
-    captions = json.load(open(args.captions))
-    sequences = sorted(captions.keys())
-    args.output_dir.mkdir(parents=True, exist_ok=True)
-    ffmpeg = ffmpeg_exe()
-
-    print(f"[reconstruct] {len(sequences)} sequences | davis-root={args.davis_root} | out={args.output_dir}")
-
-    done = skipped = failed = 0
-    missing = []
-
-    for i, seq in enumerate(sequences, 1):
-        dst = args.output_dir / f"{seq}.mp4"
-        if dst.exists() and not args.overwrite:
-            skipped += 1
-            continue
-
-        frames_dir = frames_root / seq
-        if not frames_dir.exists():
-            missing.append(seq)
-            print(f"[{i}/{len(sequences)}] MISSING  {seq}")
-            failed += 1
-            continue
-
-        try:
-            encode_sequence(ffmpeg, frames_dir, dst, args.fps)
-            print(f"[{i}/{len(sequences)}] OK  {seq}")
-            done += 1
-        except Exception as e:
-            print(f"[{i}/{len(sequences)}] ERROR  {seq}: {e}")
-            failed += 1
-
-    print(f"\n===== summary =====")
-    print(f"encoded ok       : {done}")
-    print(f"skipped (exists) : {skipped}")
-    print(f"missing upstream : {len(missing)}")
-    print(f"failed           : {failed}")
-    for seq in missing[:20]:
-        print(f"  [MISSING] {seq}")
-    if missing or failed:
-        sys.exit(1)
-
-
-if __name__ == "__main__":
-    main()

davis/videos/input_480p/bear.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ffa0652b971e27a32d5391531d97de443edbb876c3f368ed57cd17f46f043c9
+size 1516101

davis/videos/input_480p/bike-packing.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d383765fa39fc9e9896f946ce0c6383dcdca36b38833787992172a5293618809
+size 1293020

davis/videos/input_480p/blackswan.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd19ef041d79df77ee9d6b900ee9fdf803056b4a867e938afc4c8ba1438a4166
+size 1216200

davis/videos/input_480p/bmx-bumps.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:569fbda8d26cfa4a652a814fcfd20056835531bce83d6d583479c3e19a6938bb
+size 1384831

davis/videos/input_480p/bmx-trees.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c1c49d13332564fcb7aa08b9816b8a07bc168d7e2e744d7c33d27d56b9475cf
+size 2209590

davis/videos/input_480p/boat.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:655058b743cd85ed7971c03a08f2143f0d6046595de52725d2d100281ff2bdba
+size 1080880

davis/videos/input_480p/boxing-fisheye.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1469bd766511ea13d887aadb2c6ac57cd74effb56ce29e6281dbcb9903eb87cb
+size 1310456

davis/videos/input_480p/breakdance-flare.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0abc2a081e056921df79577b5ea5d9dfe5e3da7555baf663422b23dffdd5ea10
+size 1289603

davis/videos/input_480p/breakdance.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd74c19417120c776c8206650fc39e57fbcc86a79b731d782831c402ef898cbc
+size 1759013

davis/videos/input_480p/bus.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3913e96955b2f334963349c36a0b8652089f1e2936eae4661691f81ff534943
+size 1285651

davis/videos/input_480p/camel.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:105176bfbe39814c926bd62b831a38e6756baa79915f0240f86c527a5660be9b
+size 1608820

davis/videos/input_480p/car-roundabout.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:accaa99007ba69e3c27c651ac663e94b9b7c8ea54cb73d0f4ac2248aa031fd7d
+size 1554030

davis/videos/input_480p/car-shadow.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d310ee3668a46daf1dd72cc97d19911b42eaa01e708266e250a74693113f0fb2
+size 772426

davis/videos/input_480p/car-turn.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b68d64b1cdbf338b5a2e7232e938f31c650857736f016830de6a082cefad1f3
+size 1215241

davis/videos/input_480p/cat-girl.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a55a6adf05e4cd0e5cbb5bba9eebe3c07b7e902d772efcf0f6ab2dff906398a8
+size 1805372

davis/videos/input_480p/classic-car.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47d1b6ece158f3cf829eeb30aa72489da4d2372a373507c7d2bd26e186e62509
+size 1248412

davis/videos/input_480p/color-run.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6ffe96f589883d39ffd03bf8c1b45dfab84cbaaac4d9acbfc7e7a41d9d81126
+size 1082627

davis/videos/input_480p/cows.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9978cb3775670088adc4dc2aec35adfaa3ca7cf9f011fba8dad4d4c3d7ae8018
+size 2556129

davis/videos/input_480p/crossing.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56a3fcd77abf33ba529c354c6b0d96efc2cbc879c561dff3b15321fd5b4c3b9e
+size 652354

davis/videos/input_480p/dance-jump.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83c6c845e5c95633baecb873ae89df4ecd0f61cf7f40f14e58b48ee612cd71ca
+size 1134909

davis/videos/input_480p/dance-twirl.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6a0e0a1a1470df024426dde4aa7fd5d9b0f909a3b8b13f118e58e5feb784380
+size 2327661

davis/videos/input_480p/dancing.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a685eea9c7edcd2d5558b9c8c0a8999899fde0a4a258a63381102f1841d86cc7
+size 1538059

davis/videos/input_480p/disc-jockey.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d25b0b9da2649edac64438b62e7be99cfefb28edaac43ea5cf44d98ab204ce7
+size 683811

davis/videos/input_480p/dog-agility.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa0be2ba57ff1e530ce3c1b956101f8f0ac2a6d9774c3eea764bb6605fae9f90
+size 442836

davis/videos/input_480p/dog-gooses.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e4534de8258681a6bfbca45b1906bef1d430bc7609cbb6158e86d610d6e2655
+size 1309170

davis/videos/input_480p/dog.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f73fae43627e745875da89c47a8c6e534ebbb9ed85541019c896144d3ec02543
+size 2047783

davis/videos/input_480p/dogs-jump.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b13510283c14e358a00bfc3b48a366867950260b50e2da15560905c4825682f5
+size 731329

davis/videos/input_480p/dogs-scale.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40933b12b7a0c6ad1c5172519258269fae3cce840715559dddf67ba45c0a8526
+size 1220655

davis/videos/input_480p/drift-chicane.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1999f7f578dcc0072870a827b27b565aa0bb2287b146e3af2ef9292bf41aec1
+size 501965

davis/videos/input_480p/drift-straight.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c68eaed5cf32234364bee13163197272b0ce9a384ef29cac24f69a46201d95aa
+size 1097881

davis/videos/input_480p/drift-turn.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09177d06a93a1637cd13372fd740b30b4b4bfd3c34c1570fd8e53534c93cc5cc
+size 1109884

davis/videos/input_480p/drone.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ecc560521fd0c085389a21a74a45476940b763b92ca2e1df035dedcf1908385
+size 1210188

davis/videos/input_480p/elephant.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d76779253a452e7cea3364b3d8beb168c306aa619ab615e3385c76ea8bd43da
+size 1149867

davis/videos/input_480p/flamingo.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d9a7b66e124092dc29bd4f1ba7ac7290a4ed5bf19eb2d68bb3e6237896ffaba
+size 1537470

davis/videos/input_480p/goat.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e35e94dc4a117f009d3538db492d4f3776e15e5100f85c735a523210ae54280e
+size 2062781

davis/videos/input_480p/gold-fish.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c82395e6adbc85d42320ae062010b0ee301d7494d126f52cde7bdcd5dacf7f53
+size 1333596

davis/videos/input_480p/hike.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53ad72227d37f271cecdceca59b1106607d79fa52cdbe070dca44cd70b08f7ef
+size 1665971

davis/videos/input_480p/hockey.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:050c2bdad02fdfa98701d5dfa91ff8daae53dc2a303f4f20b1976c481e1e9003
+size 864860

davis/videos/input_480p/horsejump-high.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0905930261bd7d28a76ed95f1bc021ed32a58123a284144ef0519c34779b7021
+size 974606

davis/videos/input_480p/horsejump-low.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67d05a1d707f558ba2b44b8b4e1b50f541e65af61fba8d458e57d2fefa6a92f3
+size 1285747

davis/videos/input_480p/india.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af61d49a4a662d5cbbe8ecbc2b4e0d4d8cee070cfa5c83aaa8364bc57a68764a
+size 1463271

davis/videos/input_480p/judo.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc6673d25cb61cbf4e6b0e7ef6b4e5e8bdac576496ae140d98e5b81b56b8f98c
+size 276797

davis/videos/input_480p/kid-football.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c151ea93f165857611f0c07d290856b9d2b68fcb7f749767dca38be999b759e
+size 899652

davis/videos/input_480p/kite-surf.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7acf39889f7122dac98e68706be3cbe27152d1fd9cbc232a25f58f7ccab0320
+size 1480810

davis/videos/input_480p/kite-walk.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb518ced800cd995f5bcec729c6e8ee155be72ddc3d4d2adf5a95c4e743056a7
+size 1080913

davis/videos/input_480p/koala.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d07baefebec576facd309e80f9758d6b264083c261b756b654c1e8920f227e31
+size 1268473

davis/videos/input_480p/lab-coat.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44d5ddb3eb0247590b6c19281d6ab8e891bedbabe18240b2c932adfda5182bcf
+size 1197529

davis/videos/input_480p/lady-running.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71c2294a481f48ff0d7ab008e9baf6bd5576fa2797bfc95fd98e0ce3acfd7bd5
+size 838166