Add fixed Short-Horizon OVON tasks and official-source preparation guide

README.md

@@ -28,6 +28,7 @@ Language-Conditioned Local Navigation**.
 | Internet navigation | 18,184 segments | 65 segments | `annotations/internet.jsonl.gz` |
 | Simulated navigation | 2,460 episodes | 140 episodes | `annotations/simulation.jsonl.gz`, `simulation.npz` |
 | Habitat-GS-LN evaluation | 157 tasks in 45 scenes; includes a 10-task demo in 4 scenes | — | `annotations/evaluation/habitatgs_ln.json` |
+| Short-Horizon OVON evaluation | 524 tasks in 36 scenes | — | `annotations/evaluation/short_horizon_ovon.json` |
 
 The Internet collection references 13,647 upstream clips. Simulation uses
 618 training and 32 validation scenes with no scene overlap. The 45 evaluation
@@ -44,7 +45,7 @@ format without pickle; JSONL records supply their offsets and camera settings.
 From the ScaleNav code repository:
 
 ```bash
-hf download Yangyihui/ScaleNav-Data --repo-type dataset \
+hf download Yangyihui/ScaleNav-Data --repo-type dataset --revision v1.1 \
   --include 'annotations/*' --local-dir .
 ```
 
@@ -55,10 +56,17 @@ Follow the [data guide](https://github.com/yangyihui0502/ScaleNav/blob/main/docs
 to obtain upstream videos/scenes and reconstruct training inputs. Start with
 `--subset smoke` for a small training check. The separate
 [evaluation guide](https://github.com/yangyihui0502/ScaleNav/blob/main/docs/EVALUATION.md)
-covers the 10-case navigation demo and full evaluation. The full Habitat-GS-LN
-benchmark requires all 45 licensed scenes; only six are available in the pinned
-public Habitat-GS asset collection. Short-Horizon OVON assets and its filtered
-episode manifest are separate inputs, outside this annotation package.
+covers the 10-case navigation demo and both full test sets. Short-Horizon OVON
+preparation downloads official task sources, joins our fixed 524-task selection
+and builds target footprints and reference paths from locally obtained HM3D
+scenes. The selected scene assets occupy 3.48 GB.
+
+Habitat-GS-LN preparation reconstructs 39 scene visuals from official InteriorGS
+compressed files and checks that they match the benchmark assets byte for byte.
+Those scenes also require their original evaluation navmeshes (0.75 MB total),
+obtained separately from the asset authors under the upstream terms. The pinned
+Habitat-GS collection supplies the other six scenes. All 45 prepared scenes total
+8.59 GB; downloaded source caches are additional.
 
 ## Sources and license
 
@@ -74,6 +82,14 @@ The code has a separate Apache-2.0 license.
   which limit use to noncommercial research/education and prohibit redistribution
   of downloaded data.
 
+- Short-Horizon OVON references [HM3D](https://github.com/matterport/habitat-matterport-3dresearch),
+  [HM3D-OVON](https://huggingface.co/datasets/nyokoyama/hm3d_ovon) and
+  [ABot starting poses](https://huggingface.co/datasets/acvlab/ABotN-Short-Horizon-OVON).
+  Obtain their assets separately under their respective terms. This package
+  includes task IDs, rewritten instructions, target instance IDs, preparation
+  settings and checksums; the preparation script derives coordinates and
+  footprints locally.
+
 No upstream videos, scene geometry, copied geometry annotations, rendered RGB
 or latent caches are included. This annotation license does not grant additional
 rights to source assets. Render observations locally using the recorded poses.

SHA256SUMS.json

@@ -4,8 +4,8 @@
     "sha256": "600ca4e25fe11762b75a97e714707fab48bb778374e92d24c6ca068791661c11"
   },
   "README.md": {
-    "bytes": 3447,
-    "sha256": "43b627c4f8566495c44dee77f1e556f4d8b682db6c77cb744d3479f4181485d9"
+    "bytes": 4520,
+    "sha256": "d2406da5474386fe7c3da8b0be7542c1ee8ba8e0dcf8100221e7365dec4d396c"
   },
   "annotations/LICENSE": {
     "bytes": 19018,
@@ -15,6 +15,10 @@
     "bytes": 155638,
     "sha256": "6ab8bf54364c0f3dce295fb07109f37a7afabe3aaa6dc996b6bb2a2e973f189b"
   },
+  "annotations/evaluation/short_horizon_ovon.json": {
+    "bytes": 160119,
+    "sha256": "edd5f75b12ef7c7be0f563f785e6b4c7aa7c98ab51d45f47b2c704caf771674b"
+  },
   "annotations/internet.jsonl.gz": {
     "bytes": 968623,
     "sha256": "60b2881e6bc34a803befb765c554cd8963ddcccf14b091f7f6351d19e6fa3560"