|
--- |
|
task_categories: |
|
- image-to-text |
|
- text-to-image |
|
language: |
|
- en |
|
pretty_name: flickr8k |
|
size_categories: |
|
- 1K<n<10K |
|
--- |
|
|
|
You must download the dataset files manually. You can visit [this](https://github.com/jbrownlee/Datasets/releases/tag/Flickr8k) page or run `download.sh` to get files. |
|
|
|
After, you can load dataset by referencing the directory: |
|
|
|
```py |
|
import datasets |
|
ds = datasets.load_dataset("atasoglu/flickr8k-dataset", data_dir="data") |
|
print(ds) |
|
``` |
|
|
|
``` |
|
DatasetDict({ |
|
train: Dataset({ |
|
features: ['image_id', 'image_path', 'captions'], |
|
num_rows: 6000 |
|
}) |
|
test: Dataset({ |
|
features: ['image_id', 'image_path', 'captions'], |
|
num_rows: 1000 |
|
}) |
|
validation: Dataset({ |
|
features: ['image_id', 'image_path', 'captions'], |
|
num_rows: 1000 |
|
}) |
|
}) |
|
``` |
|
|
|
I don't own the copyright of the images. Please [visit](https://forms.illinois.edu/sec/1713398) for more. |
