Juliette Chataigner

Office National d'Études et de Recherches Aérospatiales

Papers

Total Citations

H-Index

About

Juliette Chataigner’s research lies at the intersection of computer vision and deep learning, with a particular focus on how video data can be leveraged to enhance single-image neural network performance. Her most notable contribution, detailed in her 2017 paper “Pertinence of Video for Single Image Deep Network,” challenges the conventional wisdom that using key frames alone is sufficient for training. Instead, she demonstrates that utilizing all successive frames from a video—despite their apparent redundancy—can dramatically boost the accuracy and robustness of deep networks, especially on medium-sized datasets. This insight offers a practical, data-efficient strategy for improving model generalization without requiring larger labeled collections. While her work has garnered modest citation counts to date, its conceptual clarity and actionable methodology have made it a valuable reference for researchers exploring data augmentation and transfer learning from video streams. Chataigner’s approach exemplifies a thoughtful re-examination of common assumptions in deep learning, providing a bridge between video-based and static-image training paradigms that continues to inform efficient model design.