estimation

Definition

Inferring a quantity you cannot observe directly from data you can — depth from a single image, 3D pose from video, a density from samples. The modifier in front says what is being estimated, and the output is typically continuous rather than a discrete label, so these tasks are scored with error magnitudes instead of accuracy. Ground truth is often imperfect itself, which is why estimation benchmarks argue about their own labels more than classification ones do.

Explainers using this term

• Paper Explained: FreeFlow — Optical Flow with a Hierarchical Transformer and No Flow-Specific PartsFreeFlow: A Bias-free Hierarchical Transformer for Optical Flow Estimation
• Paper Walkthrough: Marigold V2 — Turning an Image Editor Into a State-of-the-Art Depth Estimator on One 32GB GPUMarigold V2: Revisiting Diffusion Transformers for Monocular Depth Estimation
• Embeddings from Scratch — from word2vec Intuition to Contextual EmbeddingsEfficient Estimation of Word Representations in Vector Space