---
license: mit
tags:
  - image-classification
  - pytorch
  - resnet18
  - transfer-learning
  - computer-vision
pipeline_tag: image-classification
---

# Cat vs Dog Classifier (ResNet18 Transfer Learning)

This model is a fine-tuned **ResNet18** for binary image classification: **cat** vs **dog**.

## Model Details

- **Base model:** ResNet18 (pretrained on ImageNet)
- **Framework:** PyTorch
- **Task:** Binary image classification (cat, dog)
- **Input size:** 128x128 RGB images
- **Training method:**
  1. Feature extraction — froze all layers except the final fully connected layer, trained for 5 epochs.
  2. Fine-tuning — unfroze `layer4` of ResNet18 and trained further with a lower learning rate for 5 epochs.

## Dataset

Trained on a small subset of **CIFAR-10** (cat and dog classes only), with 100 images per class for training and 50 images per class for testing.

## How to Use

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

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
class_names = {0: "cat", 1: "dog"}

transform = transforms.Compose([
    transforms.Resize((128, 128)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                          std=[0.229, 0.224, 0.225]),
])

weights_path = hf_hub_download(repo_id="billahaiml/cat-dog-resnet18", filename="cat_dog_resnet18.pth")

model = models.resnet18(weights=None)
model.fc = nn.Linear(model.fc.in_features, 2)
model.load_state_dict(torch.load(weights_path, map_location=device))
model = model.to(device)
model.eval()

img = Image.open("your_image.jpg").convert("RGB")
img_tensor = transform(img).unsqueeze(0).to(device)

with torch.no_grad():
    output = model(img_tensor)
    probs = torch.softmax(output, dim=1)[0]
    pred = class_names[torch.argmax(probs).item()]

print(f"Prediction: {pred}")
```

## Limitations

This model was trained on a very small dataset (100 images per class), so it is intended for educational/demo purposes rather than production use.