| --- |
| license: cc-by-nc-nd-4.0 |
| pipeline_tag: image-segmentation |
| tags: |
| - one-shot anomaly-detection |
| - industrial-inspection |
| - meta-learning |
| - pytorch |
| library_name: pytorch |
| language: |
| - en |
| base_model: |
| - google/efficientnet-b4 |
| --- |
| |
| # MetaUAS Model Weights |
|
|
| This repository contains pre-trained weights for the MetaUAS anomaly detection model. This repository contains the paper described in [MetaUAS: Universal Anomaly Segmentation with One-Prompt Meta-Learning](https://huggingface.co/papers/2505.09265) |
|
|
|
|
| ## Model Files |
|
|
| | File | Description | Size | |
| |------|-------------|------| |
| | `metauas-256.ckpt` | MetaUAS model (256x256 resolution) | ~85MB | |
| | `metauas-512.ckpt` | MetaUAS model (512x512 resolution) | ~85MB | |
|
|
| ## Usage |
|
|
| ```python |
| from huggingface_hub import hf_hub_download |
| |
| # Download a specific file （"metauas-256.ckpt"） from a Hugging Face repository |
| token = "您的_HF_ACCESS_TOKEN" |
| |
| file_path = hf_hub_download( |
| repo_id="csgaobb/MetaUAS", |
| filename="metauas-256.ckpt", |
| repo_type="model" # Optional: defaults to "model", |
| token=token # 必须传入 Token |
| ) |
| |
| # Output the local cache path where the file is stored |
| print(f"File successfully downloaded to: {file_path}") |
| |
| ``` |
|
|
| ## License |
|
|
| cc-by-nc-nd-4.0 |
