Hindi VQA is a dataset for visual question answering in Hindi. It was filtered to be more balanced and processed to create sentence embeddings using a pre-trained transformer model, followed by KMeans clustering and t-SNE for visualization. The dataset was uploaded by damerajee to Hugging Face on June 2, 2024.
Use Cases
- Training visual question answering models based on Hindi-language image-text pairs.
- Analyzing clusters of similar answers based on sentence embeddings.
- Visualizing the distribution of answer types using dimensionality reduction techniques.
- Benchmarking model performance on a balanced Hindi VQA dataset.
Strengths
- The dataset was explicitly filtered to be more balanced.
- Sentence embeddings were generated using a pre-trained transformer model.
- Clustering and visualization techniques (KMeans and t-SNE) were applied to the processed data.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- huggingface
- Collection Method
- Filtered and processed from an original dataset; embeddings generated with a pre-trained transformer model.
- Freshness
- Last updated 2024-06-02 07:54:06; freshness should be verified.