Dataset Viewer
Auto-converted to Parquet Duplicate
objectID
int64
embedding
list
model
string
dim
int32
34
[ 0.005130688659846783, 0.0018449067138135433, 0.03329966217279434, 0.021693214774131775, -0.034105334430933, 0.04504737630486488, -0.00454078009352088, -0.04554551839828491, 0.007425097282975912, -0.002896029967814684, 0.0017570636700838804, 0.002548291813582182, -0.00811859592795372, 0.027...
google/siglip2-so400m-patch16-naflex
1,152
35
[ 0.002404336351901293, -0.005104158539324999, 0.02051590569317341, 0.021022023633122444, 0.0002734659647103399, -0.0041047800332307816, -0.010934400372207165, -0.05537363886833191, -0.024426626041531563, -0.029455997049808502, -0.022639835253357887, -0.01868709921836853, -0.02660229802131653,...
google/siglip2-so400m-patch16-naflex
1,152
37
[ 0.014342314563691616, -0.005803319625556469, 0.019716203212738037, 0.01112497691065073, -0.01586393639445305, 0.012873479165136814, -0.03255832567811012, -0.05286579206585884, 0.026675350964069366, -0.004997428506612778, -0.012391074560582638, 0.00030971571686677635, 0.0014419154031202197, ...
google/siglip2-so400m-patch16-naflex
1,152
38
[ 0.029771126806735992, 0.0032655561808496714, 0.034441933035850525, 0.006291836965829134, -0.006508705206215382, 0.014998350292444229, -0.03315143287181854, -0.06205962225794792, 0.016004059463739395, -0.013372767716646194, -0.008721962571144104, 0.004390249028801918, -0.0008224803023040295, ...
google/siglip2-so400m-patch16-naflex
1,152
39
[ -0.008534097112715244, -0.007881637662649155, 0.02751763164997101, 0.01961069367825985, -0.023017985746264458, -0.011585119180381298, 0.03203153237700462, -0.05231168493628502, -0.009044461883604527, -0.030277224257588387, -0.030875613912940025, -0.034206002950668335, 0.046797070652246475, ...
google/siglip2-so400m-patch16-naflex
1,152
40
[ 0.03033110499382019, 0.005426880903542042, 0.018369458615779877, 0.011002136394381523, -0.03940509259700775, 0.002591310767456889, -0.030480964109301567, -0.06259843707084656, -0.006534794345498085, -0.01013291534036398, -0.00035268577630631626, -0.011611484922468662, -0.02051123045384884, ...
google/siglip2-so400m-patch16-naflex
1,152
41
[ 0.028825340792536736, -0.004412636626511812, 0.010179968550801277, -0.00218245224095881, -0.033007506281137466, 0.00015186284144874662, -0.04153585061430931, -0.05101238191127777, -0.01749148964881897, -0.004087349865585566, -0.0016116804908961058, 0.0027525059413164854, -0.03097848594188690...
google/siglip2-so400m-patch16-naflex
1,152
42
[ -0.016106825321912766, -0.0020632226951420307, 0.003605248173698783, -0.01014593057334423, -0.02209089882671833, -0.009807241149246693, -0.032319508492946625, -0.060892876237630844, -0.0015550252282992005, -0.01521286740899086, -0.01488377247005701, -0.012428438290953636, -0.0563375279307365...
google/siglip2-so400m-patch16-naflex
1,152
43
[-0.01610684208571911,-0.002063171938061714,0.003605178091675043,-0.010145935229957104,-0.0220908150(...TRUNCATED)
google/siglip2-so400m-patch16-naflex
1,152
44
[0.0006635655299760401,-0.011909408494830132,0.005214465316385031,-0.004063892178237438,-0.017925109(...TRUNCATED)
google/siglip2-so400m-patch16-naflex
1,152
End of preview. Expand in Data Studio

metmuseum/openaccess-embeddings-siglip2-naflex

Image embeddings for every public-domain artwork in metmuseum/openaccess, produced by google/siglip2-so400m-patch16-naflex.

Column Type Notes
objectID int64 Primary key — matches objectID in metmuseum/openaccess
embedding list<float32> L2-normalised, dim = 1152
model string Source model id
dim int32 Embedding dimension

Image bytes are not stored here; join against the main dataset to recover them.

Embedding spec: dim=1152, expected image size=256px.

Joining with the main dataset

from datasets import load_dataset

meta = load_dataset("metmuseum/openaccess", split="train")
emb  = load_dataset("metmuseum/openaccess-embeddings-siglip2-naflex", split="train")

# Build an objectID -> embedding lookup, then attach to the metadata rows.
lookup = {r["objectID"]: r["embedding"] for r in emb}
joined = meta.map(lambda r: {"embedding": lookup.get(r["objectID"])})
print(joined[0].keys())

Nearest-neighbour example

import numpy as np
from datasets import load_dataset

emb = load_dataset("metmuseum/openaccess-embeddings-siglip2-naflex", split="train")
ids = np.array(emb["objectID"])
mat = np.array(emb["embedding"], dtype=np.float32)  # already L2-normalised

query = mat[0]
scores = mat @ query
top = np.argsort(-scores)[:10]
print(list(zip(ids[top].tolist(), scores[top].tolist())))

Generated by et-openaccess-embeddings.

Downloads last month
27