Muhammad-Ikhwan-Fathulloh/aksara-sunda-classification

Aksara Sunda Character Classification Dataset

This dataset is designed for the classification of Aksara Sunda characters. It was derived from an object detection dataset originally sourced from Roboflow and processed for image classification tasks.

Dataset Source

The original YOLO-formatted dataset was obtained from Roboflow: Roboflow URL: https://universe.roboflow.com/aksarasunda/aksara-sunda-eayhq/dataset/2

Dataset Structure

The dataset contains the following splits:

• train: Training set
• validation: Validation set
• test: Test set

Each sample in the dataset has the following features:

• pixel_values: A PyTorch tensor representing the image pixels (preprocessed for a Vision Transformer).
• label: An integer representing the class ID of the Aksara Sunda character.

Class Labels

The dataset contains the following Aksara Sunda characters (mapped from integer IDs to names):

• a
• ae
• ba
• ca
• da
• e
• eu
• fa
• ga
• ha
• i
• ja
• ka
• la
• ma
• na
• nga
• nya
• o
• pa
• qa
• ra
• sa
• ta
• u
• va
• wa
• xa
• ya
• za

How to Use

You can easily load this dataset from the Hugging Face Hub using the datasets library:

from datasets import load_dataset

# Load the dataset
repo_id = "Muhammad-Ikhwan-Fathulloh/aksara-sunda-classification"
loaded_dataset = load_dataset(repo_id)

print(loaded_dataset)

# Example: Access a sample from the training set
sample = loaded_dataset['train'][0]
# The label is an integer, use int2str to get the human-readable name
print(f"Sample pixel values shape: {sample['pixel_values'].shape}")
print(f"Sample label: {loaded_dataset['train'].features['label'].int2str(sample['label'])}")

If the dataset is private, you will need to authenticate with a Hugging Face token:

from huggingface_hub import login
# login(token="YOUR_HF_READ_TOKEN") # or set HF_TOKEN environment variable
# loaded_dataset = load_dataset("{repo_id}")