Datasets:
Record visual conversion provenance and exclusions
Browse files- provenance.json +55 -0
provenance.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"source": "jxu124/refcocog",
|
| 3 |
+
"revision": "55319436ad54b9480cefda6b9d64397de92456dd",
|
| 4 |
+
"format_version": 1,
|
| 5 |
+
"splits": {
|
| 6 |
+
"train": {
|
| 7 |
+
"rows": 8526,
|
| 8 |
+
"questions": 100000
|
| 9 |
+
},
|
| 10 |
+
"validation": {
|
| 11 |
+
"rows": 2566,
|
| 12 |
+
"questions": 24527
|
| 13 |
+
},
|
| 14 |
+
"test": {
|
| 15 |
+
"rows": 3980,
|
| 16 |
+
"questions": 30903
|
| 17 |
+
}
|
| 18 |
+
},
|
| 19 |
+
"conversion_sha256": "cd758bd4b55eb37a16764edd415c1f46447b28f5ee3e5692142c9b91fdbadc58",
|
| 20 |
+
"excluded_rows": 1726,
|
| 21 |
+
"license": "other",
|
| 22 |
+
"license_url": "https://cocodataset.org/#termsofuse",
|
| 23 |
+
"mirror": "tasksource/refcoco-grounding",
|
| 24 |
+
"sources": {
|
| 25 |
+
"jxu124/refcocog": "55319436ad54b9480cefda6b9d64397de92456dd",
|
| 26 |
+
"jxu124/refcoco": "9fba8200c5326e996f789191f095bd464ef1d09e",
|
| 27 |
+
"jxu124/refcocoplus": "12e20b0b6039fbf656e89e2f26597e84c1037847",
|
| 28 |
+
"Untitled111/refcoco-train2014-assets": "5daa32cfe87ea355fab400b5ec5a8a9bb476ffd2"
|
| 29 |
+
},
|
| 30 |
+
"max_train_questions": 100000,
|
| 31 |
+
"seed": 0,
|
| 32 |
+
"selection": "stable hash order of source image groups; cap counts QAs, not images",
|
| 33 |
+
"excluded_by_reason": {
|
| 34 |
+
"cross_variant_heldout_image": 1453,
|
| 35 |
+
"no_distinct_native_distractor": 25,
|
| 36 |
+
"ambiguous_duplicate_instruction": 248
|
| 37 |
+
},
|
| 38 |
+
"conversion_file_sha256": "685d07a04009ac6eac26950ba136702f2843e748e823c3d867e3237b35cf585c",
|
| 39 |
+
"parquet_sha256": {
|
| 40 |
+
"test.parquet": "88a2a24e44d2fb3060f6b11c09790c09c5721f9ef166862919fa0f2d7259ef6f",
|
| 41 |
+
"train.parquet": "273ae90ed52844595d30121d5b526c132f075d2789ac63b6dcb70781f25b335e",
|
| 42 |
+
"validation.parquet": "7e1fe337f9a3c592008536c7a93d237e1d81c0e7419b53509cd54ddc804fd06b"
|
| 43 |
+
},
|
| 44 |
+
"representation": "canonical grouped MC; original images, native candidate boxes, no synthetic negatives",
|
| 45 |
+
"choice_config": {
|
| 46 |
+
"max_options": 8,
|
| 47 |
+
"min_center_distance": 0.03,
|
| 48 |
+
"units": "0\u20131000"
|
| 49 |
+
},
|
| 50 |
+
"annotation_archive": {
|
| 51 |
+
"url": "https://s3.amazonaws.com/images.cocodataset.org/annotations/annotations_trainval2014.zip",
|
| 52 |
+
"sha256": "031296bbc80c45a1d1f76bf9a90ead27e94e99ec629208449507a4917a3bf009"
|
| 53 |
+
},
|
| 54 |
+
"split_policy": "native test subsets merged; held-out images reserved by both COCO ID and encoded-image SHA256 across variants"
|
| 55 |
+
}
|