Researchers have developed X-Lens, a novel feed-forward model for real-time metric depth estimation using a variable number of heterogeneous cameras, including fisheye and pinhole views. The model utilizes learnable calibration tokens and a Jacobian-parameterized distortion bias to achieve cross-camera consistency and robust generalization with a compact 0.04B parameters, operating at up to 41 FPS. X-Lens was trained on public datasets and the newly released OmniScene synthetic dataset, demonstrating superior accuracy and efficiency compared to existing methods. AI
IMPACT This model could improve real-time perception systems in robotics and autonomous vehicles by enabling more accurate depth estimation from diverse camera inputs.
RANK_REASON The cluster describes a research paper detailing a new model for computer vision tasks.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
