diff --git a/README.md b/README.md
index a103d80f..35178b90 100644
--- a/README.md
+++ b/README.md
@@ -71,6 +71,8 @@ We have a number of precomputed data sets in HDF5 format. All data sets have bee
 | [NYTimes](https://archive.ics.uci.edu/ml/datasets/bag+of+words)   |        256 |    290,000 |    10,000 |       100 | Angular   | [HDF5](http://ann-benchmarks.com/nytimes-256-angular.hdf5) (301MB)         |
 | [SIFT](http://corpus-texmex.irisa.fr/)                           |        128 |  1,000,000 |    10,000 |       100 | Euclidean | [HDF5](http://ann-benchmarks.com/sift-128-euclidean.hdf5) (501MB)          |
 | [Last.fm](https://github.com/erikbern/ann-benchmarks/pull/91)     |         65 |    292,385 |    50,000 |       100 | Angular   | [HDF5](http://ann-benchmarks.com/lastfm-64-dot.hdf5) (135MB)               |
+| [COCO-I2I](https://cocodataset.org/)                              |        512 |    113,287 |    10,000 |       100 | Angular   | [HDF5](https://github.com/fabiocarrara/str-encoders/releases/download/v0.1.3/coco-i2i-512-angular.hdf5) (136MB) |
+| [COCO-T2I](https://cocodataset.org/)                              |        512 |    113,287 |    10,000 |       100 | Angular   | [HDF5](https://github.com/fabiocarrara/str-encoders/releases/download/v0.1.3/coco-t2i-512-angular.hdf5) (136MB) |
 
 Results
 =======
diff --git a/ann_benchmarks/datasets.py b/ann_benchmarks/datasets.py
index 44eb8a46..8b327878 100644
--- a/ann_benchmarks/datasets.py
+++ b/ann_benchmarks/datasets.py
@@ -577,6 +577,25 @@ def dbpedia_entities_openai_1M(out_fn, n = None):
 
     write_output(X_train, X_test, out_fn, "angular")
 
+def coco(out_fn: str, kind: str):
+    assert kind in ('t2i', 'i2i')
+
+    local_fn = "coco-clip-b16-512-features.hdf5"
+    url = "https://github.com/fabiocarrara/str-encoders/releases/download/v0.1.3/%s" % local_fn
+    download(url, local_fn)
+
+    with h5py.File(local_fn, "r") as f:
+        img_X = f['img_feats'][:]
+
+        X_train, X_test = train_test_split(img_X, test_size=10_000)
+
+        if kind == 't2i':
+            # there are 5 captions per image, take the first one
+            txt_X = f['txt_feats'][::5]
+            _, X_test = train_test_split(txt_X, test_size=10_000)
+
+    write_output(X_train, X_test, out_fn, "angular")
+
 
 DATASETS: Dict[str, Callable[[str], None]] = {
     "deep-image-96-angular": deep_image,
@@ -606,6 +625,8 @@ def dbpedia_entities_openai_1M(out_fn, n = None):
     "movielens1m-jaccard": movielens1m,
     "movielens10m-jaccard": movielens10m,
     "movielens20m-jaccard": movielens20m,
+    "coco-i2i-512-angular": lambda out_fn: coco(out_fn, "i2i"),
+    "coco-t2i-512-angular": lambda out_fn: coco(out_fn, "t2i"),
 }
 
 DATASETS.update({