Flickr8K Image Features with Arabic and English Captions

Offering image features extracted from the Flickr8k dataset using a ResNeXt-152 C4 architecture. It includes Arabic and English captions and splits provided by ElJundi et al., intended for use with the OSCAR learning method.

Use Cases

• Train multimodal models using the provided image features and Arabic labels for Arabic image captioning tasks.
• Fine-tune cross-lingual models by aligning English labels with Arabic labels for translation or retrieval applications.
• Benchmark the OSCAR learning method using the pre-extracted image features and corresponding caption files.
• Analyze the alignment between object tags from the YAML configuration files and the provided image features.

Strengths

• Features are extracted using a specific, documented ResNeXt-152 C4 architecture.
• Includes parallel Arabic and English captions and splits, enabling cross-lingual analysis.
• Dataset was last updated in June 2022.

Limitations

• The number of images, rows of features, and feature dimensions are unknown.
• The dataset's license and terms of use are unspecified.
• The quality and coverage of the Arabic captions relative to the original English ones are not detailed.

Provenance

Source

Hugging Face user 'jontooy', based on the Flickr8k dataset.

Collection Method

Image features extracted using a ResNeXt-152 C4 architecture. Arabic captions and splits provided by ElJundi et al.

Freshness

Last updated 2022-06-06.