Cat and Dog Sketch Classifier

Model Details

• Model Type: Convolutional Neural Network (CNN)
• Training Data: Quick, Draw! dataset (cat and dog sketches)
• License: MIT License
• Supported Tasks: Image Classification

Usage

1. Load the Model:
   python

   1import torch 2from model import SimpleCNN 3model = SimpleCNN() 4model.load_state_dict(torch.load('cat_dog_classifier.bin')) 5model.eval()
2. Predict an Image:
   python

   1from PIL import Image 2import numpy as np 3import torch 4 5def predict_image(model, image): 6 # Preprocess the image 7 if isinstance(image, Image.Image): 8 image = image.resize((28, 28)).convert('L') 9 image = np.array(image).astype('float32') / 255.0 10 elif isinstance(image, np.ndarray): 11 if image.shape != (28, 28): 12 image = Image.fromarray(image).resize((28, 28)).convert('L') 13 image = np.array(image).astype('float32') / 255.0 14 else: 15 raise ValueError("Image must be a PIL Image or NumPy array.") 16 17 image = image.reshape(1, 1, 28, 28) 18 image_tensor = torch.tensor(image).to(device) 19 20 # Get prediction 21 model.eval() 22 with torch.no_grad(): 23 output = model(image_tensor) 24 _, predicted = torch.max(output.data, 1) 25 return 'cat' if predicted.item() == 0 else 'dog' 26 27# Example usage 28image = Image.open('path/to/your/image.png') 29prediction = predict_image(model, image) 30print(prediction)

Training the Model

License