Degradation-Aware Lightweight Vision with Few-Shot Calibration for Micro-UAV Obstacle Perception
Degradation-Aware Lightweight Vision with Few-Shot Calibration for Micro-UAV Obstacle Perception
Hani Beirami1
1) Independent Researcher
IOI Code
:
AAHD-ABKBFB
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
Reliable visual perception on micro aerial vehicles is challenging because lightweight onboard models must operate under limited computational resources while remaining robust to variations in illumination, image quality, and scene appearance. This paper presents a degradation-aware lightweight vision framework for joint obstacle-distance estimation and risk classification using MobileNetV3-small as a shared feature extractor. The network includes two prediction branches for distance regression and three-class risk classification. Its robustness is evaluated under low illumination, Gaussian blur, Gaussian noise, and partial occlusion. The baseline model achieves a distance mean absolute error of 0.104 m and a risk-classification accuracy of 95.33% on clean synthetic images, but its performance degrades substantially under low-light and noisy conditions. A degradation-aware fine-tuning strategy is therefore introduced by applying randomized camera corruptions during training. This reduces distance error by 89.39% under low illumination and 79.77% under Gaussian noise, while restoring risk accuracy to above 95% in both conditions. To reduce the remaining simulation-to-real gap, a lightweight few-shot linear calibration stage is fitted using 24 labeled real images and evaluated on 12 unseen Crazyflie-camera images across five balanced data splits. The calibrated model achieves a mean distance error of 0.306±0.070 m and a risk accuracy of 68.33±10.87%, compared with 0.757±0.017 m and 46.67±7.45% for the uncalibrated degradation-aware model. These results indicate that degradation-aware training combined with low-cost output calibration can improve robust obstacle perception without modifying the lightweight network architecture.
Keywords : micro aerial vehicles lightweight vision obstacle perception degradation aware learning few shot calibration simulation to real transfer