Researchers have introduced UniRef-UAV, a new multimodal benchmark designed to improve the ability of unmanned aerial vehicles (UAVs) to localize targets using diverse instructions. This benchmark expands upon existing methods by supporting text-only, image-only, and combined text-image queries, and can handle scenarios with no targets, single targets, or multiple targets. The team also developed UAV-URNet, a baseline model that maps various query types into a shared space and predicts target sets, demonstrating improved performance over large multimodal models in discriminating absent targets and handling variable output cardinalities. AI
IMPACT This benchmark could advance the capabilities of autonomous systems in complex visual environments, enabling more precise target identification and interaction.
RANK_REASON The cluster describes a new academic benchmark and associated model for a specific computer vision task.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
