---
license: mit
tags:
- anomaly-detection
- efficientad
- mvtec-ad
- pytorch
- computer-vision
datasets:
- mvtec_ad
metrics:
- auroc
---

# EfficientAD - Bottle

Anomaly detection model trained on MVTec AD bottle dataset using EfficientAD.

## Model Details

- **Architecture**: EfficientAD (medium)
- **Dataset**: MVTec AD - bottle
- **Task**: Anomaly Detection & Localization
- **Framework**: PyTorch

## Files

- `teacher_final.pth`: Teacher network weights
- `student_final.pth`: Student network weights
- `autoencoder_final.pth`: Autoencoder network weights
- `normalization.pth`: Normalization parameters for inference

## Usage

```python
from huggingface_hub import hf_hub_download

# Download weights
teacher_path = hf_hub_download(
    repo_id="MSherbinii/efficientad-bottle",
    filename="teacher_final.pth"
)
student_path = hf_hub_download(
    repo_id="MSherbinii/efficientad-bottle",
    filename="student_final.pth"
)
autoencoder_path = hf_hub_download(
    repo_id="MSherbinii/efficientad-bottle",
    filename="autoencoder_final.pth"
)
normalization_path = hf_hub_download(
    repo_id="MSherbinii/efficientad-bottle",
    filename="normalization.pth"
)

# Load with PyTorch
import torch
teacher = torch.load(teacher_path, map_location='cpu')
student = torch.load(student_path, map_location='cpu')
autoencoder = torch.load(autoencoder_path, map_location='cpu')
```

## Citation

Based on EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies

## License

MIT