---
library_name: pytorch
tags:
- pytorch
- image-classification
- flowers
- computer-vision
pipeline_tag: image-classification
---

# 🌸 102-Flower Image Classifier — EfficientNet-B0

PyTorch image-classification model trained to recognize **102 flower categories**.

## Results

- Architecture: **EfficientNet-B0**
- Classes: **102**
- Input: **224 × 224**
- Best validation accuracy: **94.38%**
- Epochs: **3**
- Optimizer: **AdamW**
- Learning rate: **0.001**

## Files

- `checkpoint.pth` — trained model checkpoint
- `model_config.json` — model architecture metadata
- `training_config.json` — training settings and validation result
- `class_config.json` — exact class/index mappings
- `labels.txt` — labels in model-output index order
- `requirements.txt` — Python dependencies

## Use the model

```python
import torch
import torch.nn as nn
from torchvision import models, transforms
from PIL import Image

checkpoint = torch.load("checkpoint.pth", map_location="cpu")

model = models.efficientnet_b0(weights=None)
model.classifier[1] = nn.Linear(model.classifier[1].in_features, 102)

model.load_state_dict(checkpoint["model_state_dict"])
model.eval()

idx_to_class = {
    int(k): v for k, v in __import__("json").load(open("class_config.json"))["idx_to_class"].items()
}

transform = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406],
                         [0.229, 0.224, 0.225])
])

image = Image.open("flower.jpg").convert("RGB")
x = transform(image).unsqueeze(0)

with torch.no_grad():
    probabilities = torch.softmax(model(x), dim=1)
    confidence, prediction = probabilities.max(dim=1)

idx = prediction.item()
print("Prediction:", idx_to_class[idx])
print("Confidence:", f"{confidence.item()*100:.2f}%")
```

## Checkpoint contents

The checkpoint contains:
- `epoch`
- `model_state_dict`
- `optimizer_state_dict`
- `class_to_idx`

## Training

The model was trained with transfer learning using an ImageNet-pretrained EfficientNet-B0 backbone, then fine-tuned for the 102 flower classes.

## Citation / attribution

Please retain attribution to the model author when redistributing or building upon this model. Check the original dataset's license and terms before redistribution.

## License

Add the license that applies to your model and dataset before publishing.