A comprehensive multimodal benchmark designed to evaluate the perception, cognition, and planning abilities of Multimodal Large Language Models (MLLMs) in low-altitude UAV scenarios.
MM-UAVBench focuses on assessing MLLMs' performance in UAV-specific low-altitude scenarios, with three core characteristics:
Comprehensive Task Design
19 tasks across 3 capability dimensions (perception/cognition/planning)… See the full description on the dataset page:
https://huggingface.co/datasets/daisq/MM-UAVBench.
