Decoder. plain-English AI glossary

Vision Model

● Core

Neural network trained to process images and understand visual content—classify objects, detect faces, or describe scenes.

Think of it like

Teaching a kid to recognize animals in a picture instead of just reading a list of words.

Example

ResNet-50 classifies 1000 object categories from ImageNet; YOLO detects people and cars in real-time video.

How it actually works

Convolutional layers slide filters across images, learning textures and shapes. Hierarchical: early layers learn edges, middle layers learn parts, late layers learn objects. Pre-trained on ImageNet then fine-tuned. More data-hungry than text models; 100k labeled images is a large dataset.

For product teams

Opens product capabilities in computer vision: content moderation, visual search, autonomous driving features.

For engineers

Backbone (ResNet, EfficientNet, ViT) extracts image features; head (classification, detection, segmentation) performs task-specific prediction.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome