Уровень 0 · материалов: 3
В кластер входят документы, описывающие технические методы автоматического распознавания или модерации взрослого контента, и не входят документы об удалении цензуры из изображений.
Общие признаки: компьютерное зрение, нейронные сети, детектирование контента для взрослых, автоматическая модерация
Группа выше: Прикладные задачи машинного обучения
Смысл: The main idea is to explain a comprehensive, multi-modal approach to automating the detection of adult content in videos by combining motion, color, visual patterns, and audio analysis.
Inventos developed a multi-detector system called Licenzero that identifies pornographic videos by analyzing rhythmic movement, skin tones, visual keywords, and audio patterns.
Смысл: The main idea is to explain a technical approach for detecting adult content by calculating the proportion of skin-colored pixels in video frames using the YCbCr/YUV color space.
Inventos developed a pornographic content classifier that identifies skin pixels in the YCbCr color space and uses the ratio of skin-to-total pixels to determine the probability of adult content.
Смысл: The main idea is that PornHub is leveraging computer vision and neural networks to replace manual content moderation with automated tagging and recognition, thereby improving user experience and protecting intellectual property.
PornHub is deploying an AI-powered computer vision system to automatically identify actors, sexual poses, and tags across its massive video library.