Anomalib

Anomalib is a deep learning library for anomaly detection from Intel’s Open Edge Platform team. It has about 6,300 stars on GitHub and was actively developed in 2026, with version 2.7.0 the latest release covered here.

Overview

Anomalib collects the largest public set of ready-to-use anomaly detection algorithms and benchmark datasets in one library, with a strong focus on visual inspection: detecting and localizing defects in images and video. Models are implemented on PyTorch Lightning to cut boilerplate, and training, inference, and benchmarking all run through a modular Python API or CLI. Around the models, the library adds experiment management with popular loggers, hyperparameter optimization, and benchmarking tools so algorithms can be compared fairly on public or private data. Most models export to OpenVINO for accelerated inference on Intel hardware at the edge, and the newer Anomalib Studio offers a low/no-code web app (currently pre-release) that connects cameras to training and deployment pipelines. Recent releases keep adding current algorithms — the 2.7 series alone added MH-PatchCore, the retrieval-based RAD, and the few-shot FoundAD — so the library tracks the research frontier rather than freezing it.

Key Features

• Largest algorithm collection: ready-to-use implementations of state-of-the-art anomaly detection models
• Visual focus: detect and localize anomalies in images and video, with segmentation support
• Lightning-based: modular training with far less boilerplate, via a simple API or CLI
• Experiment management: track runs with popular loggers and tune hyperparameters systematically
• Edge deployment: export models to OpenVINO for accelerated inference on Intel hardware
• Benchmarking built in: compare models across public and private datasets consistently
• Anomalib Studio: a low/no-code web app (pre-release) for training and deploying models with camera inputs

💳 Pricing

Anomalib is completely free and open source under the Apache-2.0 license, including the Studio app in the same repository.

Who Is It For

Computer vision engineers working on industrial inspection, quality control, and monitoring who need proven anomaly detection algorithms plus a credible path to edge deployment.