BrightFlow: Brightness-Change-Aware Unsupervised Learning of Optical Flow

• Vydáno v: Proceedings / IEEE Workshop on Applications of Computer Vision s. 2060 - 2069
• Hlavní autoři: MARSAL, Remi, CHABOT, Florian, LOESCH, Angelique, SAHBI, Hichem
• Typ dokumentu: Konferenční příspěvek
• Jazyk: angličtina
• Vydáno: IEEE 01.01.2023
• Témata: action recognition; Algorithms: Video recognition and understanding (tracking; and algorithms (including transfer; and un-supervised learning; Brightness; Estimation; etc.; formulations; low-shot; Machine learning architectures; Optical losses; Radiometry; Reflection; self; semi; Source coding; Training
• On-line přístup: Získat plný text
• ISSN: 2642-9381
• DOI: 10.1109/WACV56688.2023.00210

Unsupervised optical flow estimation relies on the assumption that pixels characterizing the same observed object should exhibit a stable appearance across video frames. With this assumption, the long-standing principle behind flow estimation consists in optimizing a photometric loss that maximizes...