DroneIQA-VLE: Multi-Task Drone Image Quality Assessment via Vision-Language Ensemble
This work addresses the specific problem of target-aware image quality assessment for low-altitude UAV images, but the improvement is incremental as it combines existing methods with a slight performance gain (2nd place).
The authors propose DroneIQA-VLE, a multi-task framework for drone image quality assessment that ensembles a SigLIP2-based regression pipeline with a LoRA-adapted Qwen3.5-9B multimodal LLM, achieving 2nd place in the ICME 2026 Drone-IQA Grand Challenge.
We present DroneIQA-VLE, our solution to the ICME 2026 Drone-IQA Grand Challenge on Target-aware Image Quality Assessment for Low-altitude UAV Images. The framework jointly predicts global, target, and background quality scores by ensembling two complementary pipelines: (1) SigLIP2 vision encoders with multi-task regression heads, and (2) a LoRA-adapted Qwen3.5-9B multimodal large language model for quality score regression. The final global quality prediction is obtained by arithmetically averaging the outputs of both pipelines. Our method achieves 2nd place in the challenge, demonstrating its effectiveness. The code is available at https://github.com/sunwei925/DroneIQA-VLE.