Computer Vision

Definition

AI that can understand and analyze images and video content.

Why It Matters

Text-only AI can't help when the data is a photo, a video frame, a scanned form, or a screenshot. Computer vision is the prerequisite for everything that touches pixels, OCR, captioning, segmentation, object detection, and image generation that respects layout.

Key Points

• CNNs dominated computer vision until 2021; Vision Transformers (ViTs) now match or beat them on most classification and detection tasks.
• ImageNet Top-1 accuracy has risen from 63 % (AlexNet, 2012) to 92 % (EVA, 2023) on the same benchmark.
• Object detection (YOLO, DETR) outputs bounding boxes + class labels; instance segmentation adds per-pixel masks for each detected object.
• Multimodal LLMs combine a vision encoder (like CLIP) with an LLM decoder, that architecture is how GPT-4V and Qwen-VL read images.
• Video understanding adds a temporal axis: models either process sampled frames independently or use 3D convolutions / video transformers to reason across time.

Example

A computer-vision model can look at a photo of a parking meter and output "no parking 8am-6pm Mon-Fri." The same architecture, trained on different labels, can spot a tumor in an MRI scan or count cars in satellite imagery.

Related Terms

• Multimodal AIAI models that can process multiple types of input, text, images, audio, video.
• OCR (Optical Character Recognition)AI technology that extracts text from images, PDFs and scanned documents.
• Diffusion ModelAn AI image generation technique that starts with noise and gradually refines it into a coherent image. Used by FLUX, Stable Diffusion.

Quick Facts