# CASIA-LMC-Lab/AnomalyGPT

[AAAI 2024 Oral] AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models

Repository: https://github.com/CASIA-LMC-Lab/AnomalyGPT
Canonical: https://ross.abutalabs.com/products/anomalygpt
Homepage: https://anomalygpt.github.io
Language: Python
License: NOASSERTION
License Family: other
Last push: 2023-12-20T15:06:23+00:00

## Health v2 (maintenance only)
Score: 28/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 78
- inputs: {"age_days": 1100, "days_push": 987, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1133, forks 144 (observed 2026-08-28T04:03:42.895680+00:00)

## What it is
AnomalyGPT is the first Large Vision-Language Model based method for industrial anomaly detection, identifying the presence and location of anomalies in images without manual threshold tuning. It combines a pre-trained image encoder with an LLM (Vicuna) and supports multi-turn dialogue and few-shot in-context learning.

## Use cases
- detect defects in industrial product images
- find anomalies in manufacturing inspection photos without setting thresholds
- localize defective regions in images using a vision-language model
- few-shot anomaly detection with one normal sample
- chat about images to identify what is abnormal
- train a custom anomaly detection model on my own dataset

## When to choose
- you need image-level and pixel-level anomaly detection with natural-language explanations
- you want to avoid manually tuning anomaly score thresholds
- you have GPU resources and want few-shot industrial defect detection
- you want a research-grade LVLM baseline for anomaly detection on MVTec-AD style data

## When to avoid
- you need a production-ready commercial solution - the code is CC BY-NC-SA licensed for non-commercial use only
- you lack GPU hardware to run ImageBind and Vicuna checkpoints
- you need real-time high-throughput inspection on edge devices
- you need general-purpose object detection rather than anomaly detection

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, computer-vision, image-processing, llm-inference, deep-learning
- domain: computer-vision, image-processing, large-language-models, artificial-intelligence
- platform: python
- tags: anomaly-detection, vision-language-model, industrial-inspection, multimodal, few-shot-learning, research-code, aaai-2024, manufacturing, gpu, linux

## Member repositories
- CASIA-LMC-Lab/AnomalyGPT (main) score 28

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:42.895680+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:37:23.673925+00:00, confidence not recorded.
  - readme: https://github.com/CASIA-LMC-Lab/AnomalyGPT (fetched 2026-08-28T04:03:42.895680+00:00, sha 5734a005ea8a)
  - homepage: https://anomalygpt.github.io (fetched 2026-08-29T12:42:11.828752+00:00, sha adaf0c67630b)
- Data as of 2026-08-30T08:39:29.467469+00:00.