Scholarlink

We introduce AirLetters, a new video dataset consisting of real-world videos of human-generated, articulated motions. Specifically, our dataset requires a vision model to predict letters that humans draw in the air. Unlike existing video datasets, accurate classification predictions for AirLetters rely critically on discerning motion patterns and on integrating long-range information in the video over time. An extensive evaluation of state-of-the-art image and video understanding models on AirLetters shows that these methods perform poorly and fall far behind a human baseline. Our work shows that, despite recent progress in end-to-end video understanding, accurate representations of complex articulated motions – a task that is trivial for humans – remains an open problem for end-to-end learning.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Electron microscopy image showing a detailed view of a surface with a textured, uneven pattern. The image highlights the fine structural details and variations in the surface morphology, typical of high-resolution electron microscopy. The grayscale tones emphasize the contrast between different areas, revealing intricate features and potential material composition differences. AirLetters A simple black and white symbol depicting a hand with three horizontal lines extending from the palm, representing a gesture or motion. The design is minimalistic and stylized. : An Open Vide A blue and white road sign with a red border, featuring a symbol of a bridge. The bridge icon is depicted in light blue against a white background, enclosed within a circular red border. The sign indicates the presence of a bridge or overpass. Dataset of Characters Drawn in the Air

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rishit Dagli,
  • Guillaume Berger,
  • Joanna Materzynska,
  • Ingo Bax,
  • Roland Memisevic

摘要

We introduce AirLetters, a new video dataset consisting of real-world videos of human-generated, articulated motions. Specifically, our dataset requires a vision model to predict letters that humans draw in the air. Unlike existing video datasets, accurate classification predictions for AirLetters rely critically on discerning motion patterns and on integrating long-range information in the video over time. An extensive evaluation of state-of-the-art image and video understanding models on AirLetters shows that these methods perform poorly and fall far behind a human baseline. Our work shows that, despite recent progress in end-to-end video understanding, accurate representations of complex articulated motions – a task that is trivial for humans – remains an open problem for end-to-end learning.