Datasets:
Languages:
English
Multilinguality:
monolingual
Size Categories:
1K<n<10K
Language Creators:
found
Annotations Creators:
crowdsourced
ArXiv:
License:
Commit
•
7eecddd
1
Parent(s):
d0fd144
Update README.md
Browse files
README.md
CHANGED
|
@@ -66,12 +66,18 @@ coco_url (string): url for downloading the image from the COCO server.
|
|
| 66 |
flickr_url (string): url for downloading the image from Flickr.
|
| 67 |
height (int64): height of image.
|
| 68 |
width (int64): width of image.
|
| 69 |
-
id (int64): ordered unique identifier for sample.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
## Dataset Creation
|
| 72 |
|
| 73 |
The CoCoCON dataset is created by a combination of machine + expert annotators who perturbed ground truth COCO annotations to create contrast sets.
|
| 74 |
|
|
|
|
|
|
|
| 75 |
### Licensing Information
|
| 76 |
CC-By 4.0
|
| 77 |
### Citation Information
|
|
|
|
| 66 |
flickr_url (string): url for downloading the image from Flickr.
|
| 67 |
height (int64): height of image.
|
| 68 |
width (int64): width of image.
|
| 69 |
+
id (int64): ordered unique identifier for sample.
|
| 70 |
+
|
| 71 |
+
### Data Splits
|
| 72 |
+
|
| 73 |
+
The CocoCON benchmark is an evaluation-only dataset. The data accessible through this link should be considered as the test split.
|
| 74 |
|
| 75 |
## Dataset Creation
|
| 76 |
|
| 77 |
The CoCoCON dataset is created by a combination of machine + expert annotators who perturbed ground truth COCO annotations to create contrast sets.
|
| 78 |
|
| 79 |
+
## Considerations for Using the Data
|
| 80 |
+
|
| 81 |
### Licensing Information
|
| 82 |
CC-By 4.0
|
| 83 |
### Citation Information
|