DroneDINO: Towards Heterogeneous Routed Mixture of Experts for Drone-based Unified Object Detection

Recently, the rapid development of low-altitude aerial applications has driven the need for drone-based unified detectors. In contrast to task-specific detectors that suffer from poor scalability across diverse scenarios, existing unified detectors leverage the Mixture-of-Experts (MoE) architecture to learn task-aware features from diverse datasets. However, the imbalanced multi-task data distribution leads to over-activation of experts for dominant tasks and under-activation for others. To enable balanced feature learning, this paper combines three detection paradigms (RGB, IR, and RGB-IR) into a unified framework termed DroneDINO. DroneDINO extends DINO by introducing heterogeneous routed MoEs that organize experts into three functional groups: shared, task-specific, and dynamic. Unlike conventional dynamic experts where the top-$k$ experts are activated for each input, the shared expert is activated for all inputs, while each task-specific expert is activated exclusively for the matching task. To ensure inputs are routed to appropriate experts and yield task-discriminative features, we propose a task-recognition auxiliary training strategy to penalize features with low task-discriminability. Experiments demonstrate the effectiveness and generalizability of DroneDINO, which consistently outperforms state-of-the-art unified and task-specific detectors across multiple drone-based detection benchmarks.

Drones are being deployed for increasingly diverse tasks, from package delivery to search and rescue. To operate safely, they must accurately detect objects on the ground—whether using standard color cameras by day, thermal infrared cameras at night, or both during transitional conditions. Training separate detection models for each scenario is inefficient and limits adaptability. A single model that handles all scenarios is therefore highly desirable. Existing unified approaches struggle because they learn unevenly from different types of data—performing well on common scenarios while neglecting rarer ones. To address this, we introduce DroneDINO, a detector that organizes its internal "experts" into three functional groups: shared experts that always contribute, task-specific experts that activate only for their designated scenario, and flexible experts that adapt as needed. We also introduce a training strategy that encourages the model to clearly distinguish between tasks, ensuring each input is processed by the appropriate experts. Experiments across multiple drone detection benchmarks show that DroneDINO consistently outperforms both specialized single-task models and previous unified detectors, offering a robust and adaptable solution for real-world drone applications.