Tölvusjón

Tölvusjón, or computer vision, is a field of artificial intelligence and computer science that studies how computers can gain high-level understanding from digital images or videos. The goal is to enable machines to perceive, understand, and interpret visual information in ways similar to humans, or to extract structured data from visuals for automated decision making.

It covers tasks such as image classification, object detection, instance and semantic segmentation, optical character recognition, motion tracking, 3D reconstruction, and scene understanding. Modern approaches largely rely on machine learning, especially deep learning with convolutional neural networks and transformers, trained on large annotated data sets. Common techniques include feature extraction, supervised learning, unsupervised and self-supervised learning, and 3D vision methods like multi-view geometry and neural radiance fields.

History of the field includes early research in the 1960s and 1970s focused on basic feature extraction and edge detection, with the 1990s introducing robust, hand-crafted features such as SIFT and SURF. The 2010s brought a paradigm shift with deep neural networks, leading to breakthroughs in accuracy and new applications.

Applications of tölvusjón span autonomous driving, medical imaging, industrial inspection, video surveillance, augmented reality, robotics, agriculture, and content analysis in media. The technology is integrated into consumer devices, enterprise systems, and research platforms to automate visual tasks and enable new user experiences.

Challenges include variability in lighting, occlusions, and viewpoint changes, as well as real-time processing constraints, data bias, interpretability, and privacy concerns. Evaluation relies on standardized datasets and metrics such as accuracy, average precision (AP), intersection-over-union (IoU), and F1 scores. The field continues to evolve with ethical considerations and debates about transparency, accountability, and responsible deployment.