Add dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
This repository contains the mapping from integer id's to actual label names (in HuggingFace Transformers typically called `id2label`) for several datasets.
|
| 2 |
+
|
| 3 |
+
Current datasets include:
|
| 4 |
+
- ImageNet-1k
|
| 5 |
+
- ImageNet-22k
|
| 6 |
+
- COCO detection
|
| 7 |
+
|
| 8 |
+
You can read in a label file as follows (using the `huggingface_hub` library):
|
| 9 |
+
|
| 10 |
+
```
|
| 11 |
+
from huggingface_hub import hf_hub_url, cached_download
|
| 12 |
+
import json
|
| 13 |
+
|
| 14 |
+
REPO_ID = "datasets/huggingface/label-files"
|
| 15 |
+
FILENAME = "imagenet-22k-id2label.json"
|
| 16 |
+
id2label = json.load(open(cached_download(hf_hub_url(REPO_ID, FILENAME)), "r"))
|
| 17 |
+
id2label = {int(k):v for k,v in id2label.items()}
|
| 18 |
+
```
|
| 19 |
+
|
| 20 |
+
To add an id2label mapping for a new dataset, simply define a Python dictionary, and then save that dictionary as a JSON file, like so:
|
| 21 |
+
```
|
| 22 |
+
import json
|
| 23 |
+
|
| 24 |
+
# simple example
|
| 25 |
+
id2label = {0: 'cat', 1: 'dog'}
|
| 26 |
+
|
| 27 |
+
with open('cats-and-dogs-id2label.json', 'w') as fp:
|
| 28 |
+
json.dump(id2label, fp)
|
| 29 |
+
```
|