Update README.md
Browse files
README.md
CHANGED
|
@@ -1,22 +1,37 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
configs:
|
| 4 |
-
- config_name: default
|
| 5 |
-
data_files:
|
| 6 |
-
- split: train
|
| 7 |
-
path: data/train-*
|
| 8 |
-
dataset_info:
|
| 9 |
-
features:
|
| 10 |
-
- name: id
|
| 11 |
-
dtype: string
|
| 12 |
-
- name: caption
|
| 13 |
-
dtype: string
|
| 14 |
-
- name: image
|
| 15 |
-
dtype: image
|
| 16 |
-
splits:
|
| 17 |
-
- name: train
|
| 18 |
-
num_bytes: 271106504039.8
|
| 19 |
-
num_examples: 3016640
|
| 20 |
-
download_size: 270727124231
|
| 21 |
-
dataset_size: 271106504039.8
|
| 22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
configs:
|
| 4 |
+
- config_name: default
|
| 5 |
+
data_files:
|
| 6 |
+
- split: train
|
| 7 |
+
path: data/train-*
|
| 8 |
+
dataset_info:
|
| 9 |
+
features:
|
| 10 |
+
- name: id
|
| 11 |
+
dtype: string
|
| 12 |
+
- name: caption
|
| 13 |
+
dtype: string
|
| 14 |
+
- name: image
|
| 15 |
+
dtype: image
|
| 16 |
+
splits:
|
| 17 |
+
- name: train
|
| 18 |
+
num_bytes: 271106504039.8
|
| 19 |
+
num_examples: 3016640
|
| 20 |
+
download_size: 270727124231
|
| 21 |
+
dataset_size: 271106504039.8
|
| 22 |
+
task_categories:
|
| 23 |
+
- text-to-image
|
| 24 |
+
- image-to-image
|
| 25 |
+
language:
|
| 26 |
+
- en
|
| 27 |
+
pretty_name: CC3M
|
| 28 |
+
size_categories:
|
| 29 |
+
- 1M<n<10M
|
| 30 |
+
---
|
| 31 |
+
|
| 32 |
+
This repo is CC3M's unofficial huggingface repo.
|
| 33 |
+
However, for the large picture, we process it as follow and then upload:
|
| 34 |
+
```python
|
| 35 |
+
if pil_image.width > 1024 or pil_image.height > 1024:
|
| 36 |
+
pil_image = pil_image.resize((1024, 1024), Image.BICUBIC)
|
| 37 |
+
```
|