logo left of title, CC-BY-NC-4.0, gated access, merged-model links

README.md

@@ -1,7 +1,14 @@
 ---
-license: other
-license_name: originlab-noncommercial-research
+license: cc-by-nc-4.0
 license_link: LICENSE
+extra_gated_heading: "Request access to OriginLab Game-Depth"
+extra_gated_prompt: "By requesting access you agree to the CC-BY-NC-4.0 license plus a model-release requirement: use is for non-commercial research only, and any model trained on or derived from this data must be publicly released with open weights and a model card. Commercial use requires a separate agreement with OriginLab."
+extra_gated_fields:
+  Name: text
+  Affiliation: text
+  Intended use: text
+  "I will use this dataset for non-commercial research only": checkbox
+  "I will publicly release any model I train on this dataset": checkbox
 task_categories:
 - depth-estimation
 - image-to-image
@@ -26,12 +33,12 @@ configs:
     path: data/extra-*.parquet
 ---
 
-<p align="center"><img src="figures/logo.png" width="88" alt="OriginLab"/></p>
-
-# OriginLab Game-Depth: RGB + Dense Z-Buffer Depth
+# <img src="figures/logo.png" alt="OriginLab" height="34"/> OriginLab Game-Depth: RGB + Dense Z-Buffer Depth
 
 Dense depth from game engines, as a scalable substitute for scarce real depth ground truth.
 
+**Data:** this repo (load with `load_dataset("originlab/game-depth")`). **Models / checkpoints:** [`originlab/lotus-game-depth`](https://huggingface.co/originlab/lotus-game-depth).
+
 ## Abstract
 
 Dense depth ground truth is the bottleneck in monocular depth estimation. Real sensors are sparse, noisy,
@@ -267,10 +274,7 @@ Full corpus before session-capping and the train split is 48,615 frames. Machine
 
 ## Released models
 
-Trained with this dataset and released under the same license (satisfying the model-release clause):
-
-- [`originlab/lotus-game-depth-pretrained`](https://huggingface.co/originlab/lotus-game-depth-pretrained) - pre-trained from scratch on the game depth; zero-shot KITTI AbsRel 0.191.
-- [`originlab/lotus-game-depth-nyu-ft`](https://huggingface.co/originlab/lotus-game-depth-nyu-ft) - the above fine-tuned on NYU; NYU AbsRel 0.116.
+Both models trained with this dataset are released (same license) in one repo: **[`originlab/lotus-game-depth`](https://huggingface.co/originlab/lotus-game-depth)** - the game-pretrained checkpoint (`pretrained/`, zero-shot KITTI 0.191) and the NYU fine-tuned checkpoint (`nyu-ft/`, NYU 0.116). Load with `UNet2DConditionModel.from_pretrained('originlab/lotus-game-depth', subfolder='pretrained/unet')`.
 
 ## Methodology and scope
 
@@ -292,11 +296,7 @@ is that smaller fractions imply more passes over the data (a memorization caveat
 
 ## License
 
-Released for **non-commercial research use only** under the OriginLab Game-Depth Dataset License
-([`LICENSE`](LICENSE)). Key terms: non-commercial research use, attribution required, and a
-**model-release requirement** (any model trained on or derived from this data must be publicly
-released with open weights and a model card). Commercial use requires a separate agreement with
-OriginLab.
+Released under **CC-BY-NC-4.0** ([`LICENSE`](LICENSE)) plus a **model-release requirement**: non-commercial research use, attribution, and any model trained on or derived from this data must be publicly released (open weights + model card). Commercial use requires a separate agreement with OriginLab. Access is gated - accept the terms to download.
 
 ## References