AerialDojo builds large benchmark for aerial object search tasks
AerialDojo-200K: A Large-Scale Benchmark Suite for Open-World Aerial Object-Goal Search
Computer Vision and Pattern RecognitionArtificial IntelligenceEmerging TechnologiesMultimediaRobotics
Summary
Finding objects from the air in large, open places is hard because drones must explore many kinds of environments without simple directions. The authors created AerialDojo-200K, a huge collection of simulated scenes and search tasks that covers cities, nature, and disaster areas to help train and test aerial search systems better. They carefully labeled thousands of landmarks and targets, and included video data and clear paths for each task. They also tested various advanced language-vision models and found current AI still struggles with this problem.
What this means in practice
- •For autonomous drone developers: Use AerialDojo-200K to train and benchmark drone systems for finding objects in diverse aerial environments with semantic or image-based goals.
- •For search and rescue teams: Evaluate aerial search algorithms in simulated disaster scenes to improve planning and object detection before real deployments.
Authors
Tongtong Feng, Xin Wang, Haoran Hou, Ren Wang, Weiran Wang, Shaokai Zhu, Ziqi Jia, Hao Wang, Yu-Wei Zhan, Zongyuan Wu, Jinghao Cui, Wenwu Zhu
Abstract
Open-world aerial object-goal search is a foundational yet challenging task, requiring aerial agents to autonomously explore large-scale, unstructured three-dimensional environments and reach target objects specified by semantic descriptions or reference images, rather than following route-specific instructions. However, research in this task remains at a nascent stage and relies on small, environment-specific benchmarks with heterogeneous action spaces and data formats. These limitations hinder large-scale training and cross-benchmark evaluation, constraining the scalability and generalizability of aerial agents. To address this problem, we propose AerialDojo-200K, a large-scale benchmark suite for open-world aerial object-goal search, with 3 times as many scenes and 18.7 times as many task instances as the largest existing benchmark for this task. Specifically, we construct 42 simulation scenes spanning four scene families and 21 scene types, including 18 urban, 12 natural, six infrastructure, and six disaster scenes. To ensure data quality, 12 annotators spent two months manually annotating 109 landmarks, 2099 target objects, and 2099 object anchors across these scenes. We further construct 205,732 task instances, comprising over 100K semantic-goal and over 100K image-goal instances across Base, Standard, and Long-Horizon settings. Each task instance includes a collision-free reference trajectory and corresponding multi-view video recordings. We also develop a unified evaluation framework with a scene partition comprising 21 in-distribution scenes and 21 out-of-distribution scenes. Finally, our evaluation of five open-source and four closed-source multimodal large language models reveals that there is still a long way to go toward achieving general-purpose aerial agents. All can be found at https://fengtt42.github.io/AerialDojo/.