UAV-IndoorCL: Indoor Video Dataset for Continual Learning on Drones

14,400 video frames of inter-drone and ground vehicle footage, designed to benchmark Class-Incremental Learning for resource-constrained aerial platforms. The dataset was presented in the paper 'Learning on the Fly: Replay-Based Continual Object Perception for Indoor Drones' and is hosted by author sebnae on Hugging Face. It was last updated on 2026-02-18.

Use Cases

• Benchmarking Class-Incremental Learning (CIL) algorithms based on the dataset's stated purpose
• Training object perception models for indoor drones based on the inter-drone and ground vehicle footage
• Researching continual learning for resource-constrained systems based on the dataset's design goal

Strengths

• 14,400 frames provide a substantial base for video analysis
• Specifically designed for benchmarking Class-Incremental Learning, offering a clear research focus
• Hosted on Hugging Face, which suggests a standard platform for distribution and versioning

Limitations

• Row count and file formats are unknown, which may limit 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

sebnae on Hugging Face

Collection Method

Likely captured via drone-mounted cameras, as described in the associated research paper.

Freshness

Last updated 2026-02-18 10:27:41

Geography

Indoor environments (inferred from dataset name and description)