sohumgautam/cifar_gan_model

Conditional GAN for CIFAR-10

This repository contains a trained Conditional Generative Adversarial Network (CGAN) for generating CIFAR-10 images.

Model Details

• Architecture: Conditional GAN
• Dataset: CIFAR-10
• Image Size: 32x32x3
• Latent Dimension: 100
• Number of Classes: 10

Usage

import torch
from gan_model import ConditionalGenerator, ConditionalDiscriminator

# Load the models
generator = ConditionalGenerator(latent_dim=100, num_classes=10)
discriminator = ConditionalDiscriminator(num_classes=10)

generator.load_state_dict(torch.load("generator.pth"))
discriminator.load_state_dict(torch.load("discriminator.pth"))

generator.eval()
discriminator.eval()

# Generate images
noise = torch.randn(1, 100)
label = torch.tensor([0])  # Class 0 (airplane)
fake_image = generator(noise, label)

# Evaluate with discriminator
disc_output = discriminator(fake_image, label)

Classes

0. airplane
1. automobile
2. bird
3. cat
4. deer
5. dog
6. frog
7. horse
8. ship
9. truck

Files

• generator.pth: Trained generator weights
• discriminator.pth: Trained discriminator weights
• gan_model.py: Model architecture definitions
• config.json: Model configuration