Upload README.md with huggingface_hub

README.md

@@ -28,6 +28,12 @@ Released as supporting material for an accompanying paper.
   used only for an auxiliary contrastive image-text alignment loss during training.
   It is **not used at inference time** — `forward()` only calls `image_encoder` and `cls_head`.
 
+## Dataset
+
+Dataset: [kavin-aravindhan/oct-wamd](https://huggingface.co/datasets/kavin-aravindhan/oct-wamd).
+Public access is pending IRB approval — for interim access, contact Kavin
+Aravindhan (kr3131@columbia.edu).
+
 ## Files in this repo
 
 | File | Purpose |
@@ -110,8 +116,9 @@ pip install -r requirements-train.txt
 python train.py --tfrecord-path /path/to/.tfrecord --output-dir ./runs/my_run
 ```
 
-**Data access:** the training set is clinical OCT imaging data (112 labeled images with clinical-finding captions) and is not distributed with this repo. 
-Contact the authors for access.
+**Data access:** see [Dataset](#dataset) above — the training set (112
+labeled images with clinical-finding captions) is hosted as a private HF
+dataset pending IRB approval.
 
 **Note on training-set size:** training draws 1000 samples per epoch, with replacement and augmentation, from only 112 unique images — not 1000 unique examples.