🤖 AI Summary
Existing UAV image semantic segmentation datasets suffer from limited scene diversity (predominantly urban), small scale, and low resolution, hindering model generalization and fair benchmarking. To address this, we introduce VDD—the first large-scale, high-resolution, cross-scene open-source benchmark for UAV semantic segmentation—comprising 400 densely annotated images spanning urban, industrial, rural, and natural scenes across seven semantic classes. We further unify UDD and UAVid into a comprehensive integrated dataset, IDD. This work establishes the first standardized annotation framework for non-urban aerial imagery and releases one of the largest open-source UAV segmentation datasets to date. Leveraging VDD and IDD, we conduct a systematic evaluation of seven state-of-the-art models—including SegFormer and Mask2Former—and publicly release all data, annotations, and baseline results. Our benchmark significantly enhances cross-scene generalization assessment and methodological comparability in UAV-based semantic segmentation.
📝 Abstract
Semantic segmentation of drone images is critical for various aerial vision tasks as it provides essential semantic details to understand scenes on the ground. Ensuring high accuracy of semantic segmentation models for drones requires access to diverse, large-scale, and high-resolution datasets, which are often scarce in the field of aerial image processing. While existing datasets typically focus on urban scenes and are relatively small, our Varied Drone Dataset (VDD) addresses these limitations by offering a large-scale, densely labeled collection of 400 high-resolution images spanning 7 classes. This dataset features various scenes in urban, industrial, rural, and natural areas, captured from different camera angles and under diverse lighting conditions. We also make new annotations to UDD and UAVid, integrating them under VDD annotation standards, to create the Integrated Drone Dataset (IDD). We train seven state-of-the-art models on drone datasets as baselines. It's expected that our dataset will generate considerable interest in drone image segmentation and serve as a foundation for other drone vision tasks. Datasets are publicly available at href{our website}{https://github.com/RussRobin/VDD}.
