Datasets:
Update README.md
Browse files<b>Dataset Description:</b>
This dataset, titled "Translated Flickr30k Captions - Indonesian", is derived from the original Mozilla/flickr30k-transformed-captions dataset. The primary dataset consists of image captions in English, sourced from the Flickr30k dataset, which is widely used for image captioning tasks.
<b>Translation Process:</b>
The original English captions were translated into Indonesian using the Helsinki-NLP/opus-mt-en-id translation model. This model, part of the OPUS-MT project, is designed for high-quality machine translation between English and Indonesian.
<b>Purpose:</b>
The main objective of translating the Flickr30k captions into Indonesian is to facilitate research and development in the field of image captioning within the context of the Indonesian language. By providing image captions in Indonesian, this dataset aims to support:
</ul>
<b>Image Captioning Research:</b> Enabling the creation and evaluation of image captioning models tailored for Indonesians is essential for expanding language support in AI applications.
<b>Multilingual AI Development:</b> Promoting the development of AI models that can understand and generate text in multiple languages, making AI more accessible and useful for non-English speaking communities.
<b>Dataset Enrichment:</b> Adding value to the Flickr30k dataset by providing additional linguistic diversity, which can be used for comparative studies in natural language processing (NLP) and computer vision.
</ul>
<b>Usage:</b>
Researchers and developers can use this translated dataset for training and evaluating image captioning models that generate captions in Indonesian. This can help in building applications such as:
<ul>
<b>Visual Assistive Technologies:</b> Providing image descriptions for visually impaired users in Indonesian.
<b>Multilingual Content Generation:</b> Creating diverse and inclusive content for platforms that serve Indonesian-speaking audiences.
<b>Educational Tools:</b> Developing language learning tools that integrate visual and textual information in Indonesian.
</ul>
<b>Special Condition:</b>
The image in this dataset is saved as bytes. To access the image, use the code below (scenario, accessing the first row of the dataset):
import io
dataset = load_dataset("indrad123/flickr30k-transformed-captions-indonesia")
data = pd.DataFrame({
"image": dataset["test"]["image"],
"caption": dataset["test"]["original_alt_text_id"]
})
data.loc[0, 'image']
row = data.iloc[0].squeeze()
image = Image.open(io.BytesIO(row.image['bytes'])).convert('RGB')
<b>Conclusion:</b>
The <b>"Translated Flickr30k Captions - Indonesian"</b> dataset is a valuable resource for advancing image captioning technology in the Indonesian language, fostering multilingual AI innovation, and supporting the broader goal of making AI accessible to diverse linguistic communities.