The Fréchet inception distance (FID) is a metric for assessing the quality of AI-generated images. Real and generated images are passed through a pre-trained image recognition network (Inception v3) and the statistical distributions of their features are compared. The lower the value, the closer the generated images are to real photos in realism and variety. FID was introduced in 2017 and is a standard for comparing image generators.
A generative adversarial network (GAN) consists of two neural networks trained against each other: the generator creates artificial data such as images, and the discriminator tries to tell them apart from real ones. Through this contest the results become increasingly realistic. GANs were introduced in 2014 and shaped image generation before diffusion models became more widespread; they are used, for example, for synthetic training data or image enhancement.
Stable Diffusion is a family of image generation models that create pictures from a text description. The first version was released in 2022, developed under the lead of Stability AI together with research partners. Technically it is a latent diffusion model that gradually turns image noise into a picture. Because the model weights are published, it can be run and customised locally; commercial use of newer versions is governed by Stability AI's own licence (as of October 2026).
In AI, benchmarking means systematically comparing models or systems using defined tasks and criteria. Public benchmarks measure things like general knowledge, reasoning or coding ability; for practical use, however, test cases from your own day-to-day work matter most. Typical criteria are accuracy, completeness, response time and cost. This helps decide whether a smaller, cheaper model is good enough for a task.
Image recognition is a field of computer vision in which software identifies what an image shows – objects, people, text, logos or scenes. It is usually based on neural networks trained on large numbers of labelled images. Applications range from quality control in manufacturing and sorting product photos to supporting medical imaging. With images of people, data protection and potential bias require particular attention.
