New method adapts vision models for fisheye cameras

Researchers have developed a method called Distortion Extenders (DEX) to adapt existing vision foundation models for use with fisheye cameras. These models, typically trained on standard perspective images, often produce inaccurate results when applied to fisheye images due to radial distortion. DEX introduces learnable parameters that model fisheye distortion coefficients and the distributional shift between fisheye and perspective image embeddings. By minimizing a self-supervised alignment loss, DEX transforms fisheye image embeddings to match those of perspective images, thereby recovering high-fidelity estimates for tasks like monocular depth estimation and open-vocabulary segmentation. The method is architecture-agnostic and has demonstrated improvements over baseline models on various fisheye datasets. AI

IMPACT Enables wider application of advanced vision models to data from fisheye lenses, common in robotics and surveillance.

RANK_REASON Academic paper detailing a new method for adapting computer vision models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• DagsHub
• Dex
• Distortion Extenders
• Gotit.pub
• Hugging Face
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →