Lightweight cnn models excel in uav audio classification accuracy and speed
The Unbearable Weight: Scaling Models and Methods for UAV Audio Classification
SoundMachine Learning
Summary
Distinguishing drones by their sounds is important but difficult, especially when only limited data is available. The authors compared different types of sound recognition models and ways to adapt them for drone detection. They found that smaller, simpler neural networks called lightweight CNNs work better and faster than larger transformer models for this task. Also, updating only a few model parts instead of the whole model makes retraining much quicker while keeping accuracy high.
What this means in practice
- •For uav system engineers: Implement lightweight CNN models with selective batch-norm tuning to efficiently classify drone sounds on resource-limited UAV hardware with high accuracy.
- •For security monitoring teams: Deploy efficient audio-based drone detection tools that require less computational power and retrain quickly for emerging UAV types using parameter-efficient methods.
Tested on one dataset.
Authors
Andrew P. Berg, Qian Zhang, Mia Y. Wang
Abstract
As unmanned aerial vehicles (UAVs) become increasingly prevalent in consumer and defense settings, classifying them reliably from limited, modality-specific data is an urgent challenge. The dominant approach, large pretrained networks fully fine-tuned on task data, carries a substantial computational and memory weight that is hard to bear in resource-constrained UAV deployments, where edge inference and rapid retraining for emerging platforms are both required. This paper systematically scales across both model architectures and fine-tuning methods for UAV audio classification, asking when that weight is justified and when lighter alternatives prevail. Using a custom dataset of 3,100 audio clips spanning 31 drone classes, we evaluate transformer (ViT, AST) and convolutional (custom CNN, ResNet-18/152, MobileNet-V3-S/L, EfficientNet-B0/B7) backbones under full fine-tuning, classifier-only fine-tuning, and four parameter-efficient fine-tuning (PEFT) methods: SSF, IA3, OFT, and selective batch-norm tuning. All configurations are evaluated with 5-fold cross-validation across accuracy, training time, trainable-parameter share, and inference-time memory footprint. Selective batch-norm fine-tuning of EfficientNet-B7 with three-fold augmentations achieves the highest validation accuracy (97.65% +- 0.30) while updating under 0.5% of model parameters. Across the sweep, lightweight CNNs consistently outperform transformers on both accuracy and efficiency. For UAV audio classification under data scarcity, scaling the method outperforms scaling the model.