Computer Vision

A topic-organized track for vision. Foundations first, then deep architectures, then a paper-by-paper survey of modern research areas.

Reading path

1. Foundations — image formation, filters, features, geometry. The math you need before anything is "deep".
2. Deep Vision Architectures — CNN backbones, detection, segmentation, ViT.
3. Modern Topics — ten research areas, each surveyed as a paper list:
   • Semantic Segmentation
   • Vision-Language Models
   • Neural Rendering (NeRF, 3D Gaussian Splatting)
   • Image and Video Generation
   • Geometric Deep Learning for CV
   • Representation Learning (SimCLR → DINOv2)
   • Correspondence & Structure-from-Motion
   • Safety, Robustness, Evaluation
   • Embodied CV & Robotics
   • Open-Vocabulary Detection

TIP

Vision-Language Models overlaps the LLM track — see also LLM · Multi-Modal LLMs.