Dataset Viewer
Auto-converted to Parquet Duplicate
objectID
int64
34
955k
embedding
listlengths
1.15k
1.15k
model
stringclasses
1 value
dim
int32
1.15k
1.15k
34
[ 0.017077960073947906, -0.024372605606913567, 0.02606203407049179, 0.006688839755952358, -0.011745392344892025, 0.01486252062022686, 0.022666024044156075, -0.03173312544822693, -0.009416282176971436, 0.010631573386490345, 0.001078681438229978, -0.017590031027793884, -0.020200274884700775, 0...
google/siglip2-so400m-patch14-384
1,152
35
[ 0.014886648394167423, -0.010906378738582134, 0.01709233783185482, 0.0056167361326515675, -0.012754037044942379, -0.0095239682123065, 0.0002819937944877893, -0.039537135511636734, -0.046752285212278366, -0.015796508640050888, -0.0043584490194916725, -0.029362503439188004, -0.01706008240580558...
google/siglip2-so400m-patch14-384
1,152
37
[ 0.00781311560422182, -0.000885411340277642, 0.010245701298117638, -0.010235561057925224, -0.005610446911305189, 0.01134086586534977, -0.0010999941732734442, -0.04449688643217087, -0.009370161220431328, 0.006270979065448046, 0.00022480885672848672, -0.003905159654095769, -0.011947964318096638...
google/siglip2-so400m-patch14-384
1,152
38
[ 0.01626766286790371, -0.0021343000698834658, 0.014184742234647274, -0.006452386733144522, -0.011596781201660633, 0.00690227746963501, -0.002691252389922738, -0.048583660274744034, -0.013488588854670525, -0.0003908474463969469, -0.004220329225063324, -0.01847013458609581, -0.00959395524114370...
google/siglip2-so400m-patch14-384
1,152
39
[ -0.018884174525737762, -0.009631042368710041, 0.021367672830820084, 0.003742305561900139, -0.006238492671400309, -0.00887278188019991, 0.029489940032362938, -0.039176009595394135, -0.01102713868021965, -0.005630588158965111, -0.039744291454553604, -0.015822311863303185, 0.017432378605008125,...
google/siglip2-so400m-patch14-384
1,152
40
[ 0.029029618948698044, 0.002986599924042821, -0.008507737889885902, -0.0033194448333233595, -0.040037184953689575, 0.0008215222624130547, 0.014425523579120636, -0.04036840423941612, -0.012925343587994576, -0.0010967004345729947, -0.0059319897554814816, -0.011625578626990318, -0.03623552620410...
google/siglip2-so400m-patch14-384
1,152
41
[ 0.04110347852110863, -0.005348441191017628, 0.00046067603398114443, -0.007146668620407581, -0.035747479647397995, -0.00199095718562603, 0.00016884913202375174, -0.039645999670028687, -0.02040632627904415, -0.009737040847539902, 0.007381271570920944, -0.022355923429131508, -0.0316293388605117...
google/siglip2-so400m-patch14-384
1,152
42
[ -0.018458180129528046, -0.021527621895074844, -0.019737888127565384, -0.010076159611344337, -0.012052838690578938, -0.009783169254660606, 0.006275635212659836, -0.03784715011715889, 0.006666572764515877, 0.00698036327958107, -0.010077274404466152, 0.012527834624052048, 0.004901113919913769, ...
google/siglip2-so400m-patch14-384
1,152
43
[-0.018458180129528046,-0.021527621895074844,-0.019737888127565384,-0.010076159611344337,-0.01205283(...TRUNCATED)
google/siglip2-so400m-patch14-384
1,152
44
[0.0015702027594670653,-0.02653980255126953,-0.019255293533205986,-0.004744702484458685,-0.014089972(...TRUNCATED)
google/siglip2-so400m-patch14-384
1,152
End of preview. Expand in Data Studio

metmuseum/openaccess-embeddings-siglip2

Image embeddings for every public-domain artwork in metmuseum/openaccess, produced by google/siglip2-so400m-patch14-384.

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=384px.

Joining with the main dataset

from datasets import load_dataset

meta = load_dataset("metmuseum/openaccess", split="train")
emb  = load_dataset("metmuseum/openaccess-embeddings-siglip2", 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", 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
18