Add files using upload-large-folder tool

eval_domain_surface_normals_training_v2_fixed/FINALIZATION_COMPLETE.json

@@ -0,0 +1,16 @@
+{
+  "archive_count": 4,
+  "excluded_sources": [
+    "kitti_eigen_normals",
+    "nuscenes_normals"
+  ],
+  "release": "eval_domain_surface_normals_training_v2_fixed",
+  "row_count": 83225,
+  "source_groups": [
+    "diode",
+    "nyuv2",
+    "scannetv2",
+    "vkitti2_nonclone"
+  ],
+  "status": "PASS"
+}

eval_domain_surface_normals_training_v2_fixed/MULTIPART_MANIFEST.json

@@ -0,0 +1,52 @@
+{
+  "archives": [
+    {
+      "assembly": "cat archives/diode_normals_train_v1.tar.parts/diode_normals_train_v1.tar.part-* > archives/diode_normals_train_v1.tar",
+      "logical_bytes": 47938160640,
+      "logical_path": "archives/diode_normals_train_v1.tar",
+      "logical_sha256": "eff7baa490d448b4a4a83f29e55d763398b34d78e56e0c80dbe0fbb5cd7d0d15",
+      "parts": [
+        {
+          "bytes": 17179869184,
+          "path": "diode_normals_train_v1.tar.part-0000",
+          "sha256": "7a9d77bb1bfa77640876e6bd10129d3ed7e795915ba360c1b777b5e4de9ffaa7"
+        },
+        {
+          "bytes": 17179869184,
+          "path": "diode_normals_train_v1.tar.part-0001",
+          "sha256": "1800b1f4e8dcd747364de4106cdf0d19dfe5d9a12fa0fd911af042025c388da7"
+        },
+        {
+          "bytes": 13578422272,
+          "path": "diode_normals_train_v1.tar.part-0002",
+          "sha256": "92718cde417da8cb02474036785163e547b45494170839fcd3f788991c60801f"
+        }
+      ]
+    },
+    {
+      "assembly": "cat archives/vkitti2_nonclone_normals_train_v2_fixed.tar.parts/vkitti2_nonclone_normals_train_v2_fixed.tar.part-* > archives/vkitti2_nonclone_normals_train_v2_fixed.tar",
+      "logical_bytes": 40490516480,
+      "logical_path": "archives/vkitti2_nonclone_normals_train_v2_fixed.tar",
+      "logical_sha256": "d7826a7423759e5401b45f17099cf976efffb5ae2bc8d351ff7a7b23d3073429",
+      "parts": [
+        {
+          "bytes": 17179869184,
+          "path": "vkitti2_nonclone_normals_train_v2_fixed.tar.part-0000",
+          "sha256": "d5d615f01473b5ac971aa5e001cbdcf75aa4ace6a8ab4f6819e13c4fda0dbdd4"
+        },
+        {
+          "bytes": 17179869184,
+          "path": "vkitti2_nonclone_normals_train_v2_fixed.tar.part-0001",
+          "sha256": "6b442b87e395f3fed595fbee2ffab35df694e8f8de91522b32512223ddcf4b34"
+        },
+        {
+          "bytes": 6130778112,
+          "path": "vkitti2_nonclone_normals_train_v2_fixed.tar.part-0002",
+          "sha256": "b9556052dc4768b335358f4ba2fd73c7bee17ab78ea0c663a6b2d292522f76c7"
+        }
+      ]
+    }
+  ],
+  "chunk_bytes": 17179869184,
+  "schema_version": 1
+}

eval_domain_surface_normals_training_v2_fixed/README.md

@@ -0,0 +1,9 @@
+# Eval-domain surface-normal training splits v2 (fixed)
+
+This release supersedes only the **surface-normal** portion of `eval_domain_geometry_training_v1`; the v1 depth archives are unchanged. It uses one physical convention throughout: camera-space +x image-right, +y image-up, +z toward the camera, stored as `vision_banana_x_left_red` (`R=(1-x)/2`, `G=(1+y)/2`, `B=(1+z)/2`).
+
+Corrections: NYUv2 now uses the official GeoNet normal labels distributed with DSINE for both train-domain supervision and evaluation; Virtual KITTI 2 applies its source-to-canonical transform exactly once. DIODE and ScanNet v2 retain their audited v1 normal archives. KITTI and nuScenes normal archives are deliberately excluded because their sparse masks do not survive the released hard majority-valid latent-mask policy. Their depth archives remain valid in v1.
+
+Use `configs/surface_normals_equal_source_fixed.json` with the indexed prepared-TAR backend and all flags in `configs/training_contract.json`. To download, run `python configs/download_release.py --prefix eval_domain_surface_normals_training_v2_fixed --output DATA_ROOT`. The script prints the directly usable `READY_DATA_ROOT`.
+
+The DSINE bundle and each underlying dataset retain their original licenses and terms; this release does not replace them.

eval_domain_surface_normals_training_v2_fixed/SHA256SUMS

@@ -0,0 +1,4 @@
+eff7baa490d448b4a4a83f29e55d763398b34d78e56e0c80dbe0fbb5cd7d0d15  archives/diode_normals_train_v1.tar
+52cf593ac220db4762e428f05eb81228a790d16ebfe871d2314b01cc31ee6fef  archives/nyuv2_normals_train_v2_fixed.tar
+bf195d4f4374641b9e24000415ea49bc629948aae724556a59709e10f61d78d0  archives/scannetv2_normals_train_v1.tar
+d7826a7423759e5401b45f17099cf976efffb5ae2bc8d351ff7a7b23d3073429  archives/vkitti2_nonclone_normals_train_v2_fixed.tar

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.ovb-index.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c8b7c1e58ee1daa64f0a4dedf2aef8a735245a09403500adff4db1822eb542d
+size 67198976

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.parts/diode_normals_train_v1.tar.part-0000

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a9d77bb1bfa77640876e6bd10129d3ed7e795915ba360c1b777b5e4de9ffaa7
+size 17179869184

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.parts/diode_normals_train_v1.tar.part-0001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1800b1f4e8dcd747364de4106cdf0d19dfe5d9a12fa0fd911af042025c388da7
+size 17179869184

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.parts/diode_normals_train_v1.tar.part-0002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92718cde417da8cb02474036785163e547b45494170839fcd3f788991c60801f
+size 13578422272

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.sha256

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d5c02c7a6c46caea00a05ba27162300f95ac2a2e398b499f8003bc4bc991279
+size 93

eval_domain_surface_normals_training_v2_fixed/archives/diode_normals_train_v1.tar.validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cc3db47dc238bba1440564028a83d433b665fcf690b1d30a5cbe675a182bc32
+size 1893

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52cf593ac220db4762e428f05eb81228a790d16ebfe871d2314b01cc31ee6fef
+size 486256640

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar.ovb-index.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96b6ae015d0892e537eb1be202291334c0a4de93dbe738ffddb9b17c92db91f7
+size 3432448

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar.semantic-validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a89566a8d6f61c1bc6f8cbe6f6c9ecc7b9343caf84bfffca971eeba0cc6b4f81
+size 507

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar.sha256

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9ec164cda0eff18ebf59a1cbdd0a4982d213719758f1b250e192476c26ebc86
+size 99

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar.source-oracle.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8607a3c6984f9c9c4aa0a6e63ac8a9f153876de2cd56083300fd5df52993db09
+size 624

eval_domain_surface_normals_training_v2_fixed/archives/nyuv2_normals_train_v2_fixed.tar.validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2960fbcf5468d107d39fea81fd721a1a50523c3659bcd70d22f877d212f1efc
+size 1973

eval_domain_surface_normals_training_v2_fixed/archives/scannetv2_normals_train_v1.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf195d4f4374641b9e24000415ea49bc629948aae724556a59709e10f61d78d0
+size 17182883840

eval_domain_surface_normals_training_v2_fixed/archives/scannetv2_normals_train_v1.tar.ovb-index.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb096fb2abd067144b7f851bb46ae6ea76b81ed4cd0ae152e218ac6dd53adec5
+size 51834880

eval_domain_surface_normals_training_v2_fixed/archives/scannetv2_normals_train_v1.tar.sha256

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a08dc1f8820cd1d46e3ef7671b4eb55a582d4db3a45fa7de9061ced1fe0c8d2e
+size 97

eval_domain_surface_normals_training_v2_fixed/archives/scannetv2_normals_train_v1.tar.validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4b3b8ca0894a048c9225706cdc75c0ac79b882fa3a872d6aab1ed361397dc9e
+size 1858

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.ovb-index.sqlite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f034b6c99b38e4e487e05473f27e53835060a72eb723c4a8f86f68eb01f8584
+size 185876480

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.parts/vkitti2_nonclone_normals_train_v2_fixed.tar.part-0000

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5d615f01473b5ac971aa5e001cbdcf75aa4ace6a8ab4f6819e13c4fda0dbdd4
+size 17179869184

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.parts/vkitti2_nonclone_normals_train_v2_fixed.tar.part-0001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b442b87e395f3fed595fbee2ffab35df694e8f8de91522b32512223ddcf4b34
+size 17179869184

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.parts/vkitti2_nonclone_normals_train_v2_fixed.tar.part-0002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9556052dc4768b335358f4ba2fd73c7bee17ab78ea0c663a6b2d292522f76c7
+size 6130778112

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.semantic-validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90162986c781e2c2fe3e8284dd4f859f4223e04303df1dde8e547d6bf530d85b
+size 410

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.sha256

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4265dece57650c575c202538579898a82b4304a848a708b96a1ddd661139e591
+size 110

eval_domain_surface_normals_training_v2_fixed/archives/vkitti2_nonclone_normals_train_v2_fixed.tar.validation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37d0cb49f9898de63a27f8006389183e1595ccb797d27e1c79df4f433cd9f991
+size 1922

eval_domain_surface_normals_training_v2_fixed/configs/download_release.py

@@ -0,0 +1,133 @@
+#!/usr/bin/env python3
+"""Download, verify, reconstruct, and make the Hub release directly usable."""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+import os
+import shutil
+import sqlite3
+import urllib.parse
+import urllib.request
+from pathlib import Path
+
+
+PREFIX = "eval_domain_geometry_training_v1"
+CHUNK = 16 * 1024**2
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(CHUNK), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def inventory(repo: str, revision: str, prefix: str) -> list[dict]:
+    url = (
+        f"https://huggingface.co/api/datasets/{repo}/tree/"
+        f"{urllib.parse.quote(revision, safe='')}/{urllib.parse.quote(prefix, safe='')}?recursive=true&expand=false&limit=1000"
+    )
+    with urllib.request.urlopen(url, timeout=60) as response:
+        rows = json.load(response)
+    return [row for row in rows if row.get("type") == "file"]
+
+
+def download(repo: str, revision: str, row: dict, root: Path) -> Path:
+    relative = row["path"]
+    destination = root / relative
+    partial = destination.with_suffix(destination.suffix + ".partial")
+    expected = int(row["size"])
+    destination.parent.mkdir(parents=True, exist_ok=True)
+    if destination.is_file() and destination.stat().st_size == expected:
+        return destination
+    current = partial.stat().st_size if partial.exists() else 0
+    request = urllib.request.Request(
+        f"https://huggingface.co/datasets/{repo}/resolve/{urllib.parse.quote(revision, safe='')}/"
+        f"{urllib.parse.quote(relative, safe='/')}?download=true",
+        headers={"Range": f"bytes={current}-"} if current else {},
+    )
+    with urllib.request.urlopen(request, timeout=120) as response:
+        mode = "ab" if current and getattr(response, "status", None) == 206 else "wb"
+        with partial.open(mode) as output:
+            while block := response.read(CHUNK):
+                output.write(block)
+    if partial.stat().st_size != expected:
+        raise IOError(f"size mismatch for {relative}")
+    partial.replace(destination)
+    return destination
+
+
+def copy_tree_files(source: Path, destination: Path) -> None:
+    for path in source.rglob("*"):
+        if not path.is_file() or ".parts" in path.parts:
+            continue
+        target = destination / path.relative_to(source)
+        target.parent.mkdir(parents=True, exist_ok=True)
+        if not target.exists():
+            try:
+                os.link(path, target)
+            except OSError:
+                shutil.copy2(path, target)
+
+
+def retarget_index(archive: Path, index: Path, digest: str) -> None:
+    with sqlite3.connect(index) as db:
+        row = db.execute("SELECT value FROM metadata WHERE key='archive_fingerprint'").fetchone()
+        if row is None:
+            raise ValueError(f"missing archive fingerprint: {index}")
+        fingerprint = json.loads(row[0])
+        fingerprint.update(size=archive.stat().st_size, mtime_ns=archive.stat().st_mtime_ns, sha256=digest)
+        db.execute(
+            "UPDATE metadata SET value=? WHERE key='archive_fingerprint'",
+            (json.dumps(fingerprint, sort_keys=True),),
+        )
+        db.commit()
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--repo-id", default="arpitjadon/OVB_Data")
+    parser.add_argument("--revision", default="main")
+    parser.add_argument("--prefix", default=PREFIX)
+    parser.add_argument("--output", type=Path, required=True)
+    args = parser.parse_args()
+    output = args.output.resolve()
+    downloaded = output / "download"
+    ready = output / "ready" / args.prefix
+    for row in inventory(args.repo_id, args.revision, args.prefix):
+        download(args.repo_id, args.revision, row, downloaded)
+    release = downloaded / args.prefix
+    manifest = json.loads((release / "MULTIPART_MANIFEST.json").read_text(encoding="utf-8"))
+    ready.mkdir(parents=True, exist_ok=True)
+    copy_tree_files(release, ready)
+    for entry in manifest["archives"]:
+        logical = ready / entry["logical_path"]
+        logical.parent.mkdir(parents=True, exist_ok=True)
+        partial = logical.with_suffix(logical.suffix + ".partial")
+        with partial.open("wb") as output_stream:
+            part_root = release / f"{entry['logical_path']}.parts"
+            for part in entry["parts"]:
+                with (part_root / part["path"]).open("rb") as source:
+                    shutil.copyfileobj(source, output_stream, length=CHUNK)
+        if sha256(partial) != entry["logical_sha256"]:
+            raise ValueError(f"multipart SHA-256 mismatch: {logical}")
+        partial.replace(logical)
+    sums = {}
+    for line in (ready / "SHA256SUMS").read_text(encoding="utf-8").splitlines():
+        digest, name = line.split(None, 1)
+        sums[name.strip()] = digest
+    for relative, digest in sums.items():
+        archive = ready / relative
+        if sha256(archive) != digest:
+            raise ValueError(f"archive SHA-256 mismatch: {archive}")
+        retarget_index(archive, archive.with_name(archive.name + ".ovb-index.sqlite"), digest)
+    print(f"READY_DATA_ROOT={ready}")
+    print("Use the JSON dataset specification under READY_DATA_ROOT/configs for training.")
+
+
+if __name__ == "__main__":
+    main()

eval_domain_surface_normals_training_v2_fixed/configs/surface_normals_equal_source_fixed.json

@@ -0,0 +1,35 @@
+{
+  "datasets": [
+    {
+      "index_path": "archives/diode_normals_train_v1.tar.ovb-index.sqlite",
+      "name": "diode_normals_train_v1",
+      "path": "archives/diode_normals_train_v1.tar",
+      "type": "indexed_prepared_tar",
+      "weight": 0.25
+    },
+    {
+      "index_path": "archives/nyuv2_normals_train_v2_fixed.tar.ovb-index.sqlite",
+      "name": "nyuv2_normals_train_v2_fixed",
+      "path": "archives/nyuv2_normals_train_v2_fixed.tar",
+      "type": "indexed_prepared_tar",
+      "weight": 0.25
+    },
+    {
+      "index_path": "archives/scannetv2_normals_train_v1.tar.ovb-index.sqlite",
+      "name": "scannetv2_normals_train_v1",
+      "path": "archives/scannetv2_normals_train_v1.tar",
+      "type": "indexed_prepared_tar",
+      "weight": 0.25
+    },
+    {
+      "index_path": "archives/vkitti2_nonclone_normals_train_v2_fixed.tar.ovb-index.sqlite",
+      "name": "vkitti2_nonclone_normals_train_v2_fixed",
+      "path": "archives/vkitti2_nonclone_normals_train_v2_fixed.tar",
+      "type": "indexed_prepared_tar",
+      "weight": 0.25
+    }
+  ],
+  "samples_per_epoch": 1000000,
+  "sampling": "equal mass over four dense, semantically audited source datasets",
+  "task": "surface_normal_estimation"
+}

eval_domain_surface_normals_training_v2_fixed/configs/training_contract.json

@@ -0,0 +1,15 @@
+{
+  "mask_policy": "hard majority-valid latent cells; no erosion",
+  "normal_preprocessing": "decode vectors, resize vectors, renormalize, then encode",
+  "normal_prompt_contract": "campaign_short",
+  "normal_target_encoding": "vision_banana_x_left_red",
+  "trainer_cli_required": [
+    "--normal_target_encoding vision_banana_x_left_red",
+    "--normal_target_resize_policy vector_field_renormalize_then_encode",
+    "--loss_mask_policy enabled",
+    "--loss_mask_mode hard",
+    "--loss_mask_min_valid_fraction 0.5",
+    "--loss_mask_min_sample_valid_fraction 0",
+    "--loss_mask_erosion_radius 0"
+  ]
+}

eval_domain_surface_normals_training_v2_fixed/manifests/archive_inventory.json

@@ -0,0 +1,49 @@
+{
+  "archives": [
+    {
+      "archive": "diode_normals_train_v1.tar",
+      "authoritative_source_validation": null,
+      "bytes": 47938160640,
+      "rows": 24884,
+      "semantic_validation": "audited-v1-source",
+      "sha256": "eff7baa490d448b4a4a83f29e55d763398b34d78e56e0c80dbe0fbb5cd7d0d15",
+      "source_group": "diode",
+      "task": "surface_normal_estimation",
+      "validation": "diode_normals_train_v1.tar.validation.json"
+    },
+    {
+      "archive": "nyuv2_normals_train_v2_fixed.tar",
+      "authoritative_source_validation": "nyuv2_normals_train_v2_fixed.tar.source-oracle.json",
+      "bytes": 486256640,
+      "rows": 795,
+      "semantic_validation": "nyuv2_normals_train_v2_fixed.tar.semantic-validation.json",
+      "sha256": "52cf593ac220db4762e428f05eb81228a790d16ebfe871d2314b01cc31ee6fef",
+      "source_group": "nyuv2",
+      "task": "surface_normal_estimation",
+      "validation": "nyuv2_normals_train_v2_fixed.tar.validation.json"
+    },
+    {
+      "archive": "scannetv2_normals_train_v1.tar",
+      "authoritative_source_validation": null,
+      "bytes": 17182883840,
+      "rows": 19278,
+      "semantic_validation": "audited-v1-source",
+      "sha256": "bf195d4f4374641b9e24000415ea49bc629948aae724556a59709e10f61d78d0",
+      "source_group": "scannetv2",
+      "task": "surface_normal_estimation",
+      "validation": "scannetv2_normals_train_v1.tar.validation.json"
+    },
+    {
+      "archive": "vkitti2_nonclone_normals_train_v2_fixed.tar",
+      "authoritative_source_validation": null,
+      "bytes": 40490516480,
+      "rows": 38268,
+      "semantic_validation": "vkitti2_nonclone_normals_train_v2_fixed.tar.semantic-validation.json",
+      "sha256": "d7826a7423759e5401b45f17099cf976efffb5ae2bc8d351ff7a7b23d3073429",
+      "source_group": "vkitti2_nonclone",
+      "task": "surface_normal_estimation",
+      "validation": "vkitti2_nonclone_normals_train_v2_fixed.tar.validation.json"
+    }
+  ],
+  "status": "PASS"
+}

eval_domain_surface_normals_training_v2_fixed/manifests/excluded_normal_sources.json

@@ -0,0 +1,4 @@
+{
+  "kitti_eigen_normals": "Excluded: sparse image-space supervision is almost entirely removed by the released hard majority-valid latent-mask policy.",
+  "nuscenes_normals": "Excluded: sparse image-space supervision is entirely removed in the audited 512-resolution hard majority-valid latent-mask sample."
+}