Cranfield Synthetic Drone Classification: Four-Class Aerial Imagery

Cranfield Synthetic Drone Classification is a dataset created for the paper 'Drone Model Classification Using Convolutional Neural Network Trained on Synthetic Data'. It contains synthetic images across four classes: DJI Mavic, DJI Phantom, DJI Inspire, and No Drone. The dataset was uploaded to Hugging Face by user mazqtpopx and last updated in November 2024.

Use Cases

• Training a convolutional neural network for drone model classification based on the synthetic imagery.
• Benchmarking synthetic data performance against real-world drone imagery.
• Developing a binary drone detection system based on the 'No Drone' class.
• Fine-tuning a pre-trained image classifier on the four specific DJI drone models.

Strengths

• Dataset is directly linked to a peer-reviewed academic paper published in the Journal of Imaging.
• Provides a clear, defined classification task with four distinct classes.
• Focuses on synthetic data, a controlled and reproducible data generation method.

Limitations

• Dataset size, row count, and file formats are unknown, limiting suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Paper authors: Mariusz Wisniewski, Zeeshan A. Rana, Ivan Petrunin.

Collection Method

Synthetic data generation, as described in the associated research paper.

Time Range

null

Freshness

Last updated 2024-11-12 11:04:24; freshness should be verified.

Geography

null