Deepfake Detection: 15,000 Multi-Modal Tensors for Direct Training

15,000 multi-modal tensors combine CLIP embeddings with statistical features for deepfake detection. The dataset is optimized for direct training of machine learning models. The author, organization, and last update date are unknown.

Use Cases

• Train binary classifiers for deepfake detection based on hybrid tensor features.
• Benchmark model performance on multi-modal data combining visual and statistical signals.
• Develop ensemble detection methods leveraging both CLIP embeddings and statistical features.

Strengths

• 15,000 data points provide a substantial base for model training.
• Data is pre-processed into tensors optimized for direct training, reducing preprocessing overhead.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to Kaggle.