README.md

---
library_name: lucid
license: bsd-3-clause
tags:
  - image-classification
  - inception
  - lucid
datasets:
  - imagenet-1k
pipeline_tag: image-classification
model-index:
  - name: inception-v3
    results:
      - task: { type: image-classification }
        dataset: { name: ImageNet-1K, type: imagenet-1k }
        metrics:
          - { type: acc@1, value: 77.294 }
          - { type: acc@5, value: 93.45 }
---

# Inception v3

> Szegedy et al., 2015 — *Rethinking the Inception Architecture for Computer Vision* (arXiv:1512.00567)

[Lucid](https://github.com/ChanLumerico/lucid) port of `torchvision/Inception_V3_Weights.IMAGENET1K_V1`,
converted to Lucid-native safetensors.

## Available weights

| Tag | acc@1 | acc@5 | Params | GFLOPs | Size | Source |
|---|---|---|---|---|---|---|
| `IMAGENET1K_V1` *(default)* | 77.294 | 93.45 | 23.8M | 5.713 | 91.11 MB | torchvision |

## Usage

```python
import lucid.models as models
from lucid.models.weights import InceptionV3Weights

# default tag
model = models.inception_v3_cls(pretrained=True)

# explicit tag (enum or string)
model = models.inception_v3_cls(weights=InceptionV3Weights.IMAGENET1K_V1)
model = models.inception_v3_cls(pretrained="IMAGENET1K_V1")

# preprocessing travels with the weights
weights = InceptionV3Weights.IMAGENET1K_V1
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits
```

## Conversion

Converted from `torchvision/Inception_V3_Weights.IMAGENET1K_V1` via
`python -m tools.convert_weights inception_v3 --tag IMAGENET1K_V1`.
Key mapping + numerical parity verified against the source.

## License

`bsd-3-clause` — inherited from the original weights.

## Citation

```
@inproceedings{szegedy2016rethinking,
  title={Rethinking the Inception Architecture for Computer Vision},
  author={Szegedy, Christian and Vanhoucke, Vincent and Ioffe, Sergey and Shlens, Jon and Wojna, Zbigniew},
  booktitle={CVPR}, year={2016}
}
```