An active learning framework for drone classification in radio frequency domain
Bibliographic record
Abstract
Radio-frequency–based drone classification is a critical capability for modern antidrone systems. However, the development of dependable artificial intelligence models in this domain is hindered by the high cost and complexity of expert data labeling. This challenge is particularly pronounced in radio-frequency signal analysis, where labeled datasets are typically small, partially unlabeled, and continuously evolving during system deployment. This paper proposes a tailored active learning framework for drone classification in the radio-frequency domain, integrating human expertise into an iterative learning process that selectively queries the most informative unlabeled samples. By prioritizing sample informativeness, the proposed framework aims to achieve high detection performance while significantly reducing labeling effort. Theapproach is evaluated using the VTI_USRP_DroneSET dataset, comprising radio-frequency spectrograms acquired in realistic outdoor conditions within the 2.4 GHz frequency band. Experimental results demonstrate that the proposed active learning strategy achieves mAP50–95 performance comparable to conventional supervised learning while requiring only one quarter of the labeled data. The results confirm that active learning enables data-efficient radio-frequency based drone classification without compromising detection accuracy. Furthermore, near-optimal performance is consistently obtained with an optimal training duration of 80 epochs, reducing both annotation and computational costs. These findings confirm that active learning provides a data-efficient and cost-effective solution for RF-based drone classification and is well suited for real-time deployment in operational antidrone systems where labeled data are scarce and continuously acquired.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".