---
tags:
  - image-classification
  - biology
library_name: pytorch
license: mit
---

# Speyeria Classifier

This repository contains a ResNet-50 based PyTorch classifier for identifying
Speyeria species from images.

## Labels

The model predicts the following species (short labels are used in `config.json`):

- S. adiaste
- S. aphrodite
- S. atlantis
- S. callippe
- S. carolae
- S. coronis
- S. cybele
- S. diana
- S. edwardsii
- S. egleis
- S. hesperis
- S. hydaspe
- S. idalia
- S. mormonia
- S. nokomis
- S. zerene

## Usage

Use `model.py` to instantiate the architecture and load weights:

```python
from safetensors.torch import load_file
from model import create_model

model = create_model(num_classes=16)
state_dict = load_file("model.safetensors")
model.load_state_dict(state_dict)
model.eval()
```

## Notes

- `config.json` contains `id2label`/`label2id` mappings.
- Weights are stored in `.safetensors` format for safer deserialization.