RAPTOR: Real-Time High-Resolution Video Prediction for UAVs

Published:Dec 25, 2025 15:12
1 min read
ArXiv

Analysis

This paper addresses the critical need for real-time, high-resolution video prediction in autonomous UAVs, a domain where latency is paramount. The authors introduce RAPTOR, a novel architecture designed to overcome the limitations of existing methods that struggle with speed and resolution. The core innovation, Efficient Video Attention (EVA), allows for efficient spatiotemporal modeling, enabling real-time performance on edge hardware. The paper's significance lies in its potential to improve the safety and performance of UAVs in complex environments by enabling them to anticipate future events.

Reference

RAPTOR is the first predictor to exceed 30 FPS on a Jetson AGX Orin for $512^2$ video, setting a new state-of-the-art on UAVid, KTH, and a custom high-resolution dataset in PSNR, SSIM, and LPIPS. Critically, RAPTOR boosts the mission success rate in a real-world UAV navigation task by 18%.