Mini Project: Dog vs Cat Classifier with Transfer Learning
Build a production-quality binary image classifier using EfficientNet-B0 fine-tuned on the Dogs vs Cats dataset (25,000 images). Achieves 99%+ accuracy in under 10 epochs by leveraging ImageNet pre-training — demonstrating the power of transfer learning.
Dogs vs Cats Transfer Learning Project
import torch
import torch.nn as nn
import torchvision.models as models
import torchvision.transforms as transforms
from torchvision.datasets import ImageFolder
from torch.utils.data import DataLoader, random_split
import torch.optim as optim
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
# Setup
# Dataset: kaggle datasets download -d salader/dogs-vs-cats
# Folder structure expected:
# data/dogs-vs-cats/
# train/
# cats/ (*.jpg)
# dogs/ (*.jpg)
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
TRAIN_DIR = "data/dogs-vs-cats/train"
BATCH_SIZE = 64
N_EPOCHS = 10
IMG_SIZE = 224
# Transforms
train_tfm = transforms.Compose([
transforms.Resize((IMG_SIZE + 32, IMG_SIZE + 32)),
transforms.RandomCrop(IMG_SIZE),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(15),
transforms.ColorJitter(brightness=0.2, saturation=0.2),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
val_tfm = transforms.Compose([
transforms.Resize((IMG_SIZE, IMG_SIZE)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
# Load dataset
full_ds = ImageFolder(TRAIN_DIR, transform=train_tfm)
print(f"Classes: {full_ds.classes}") # ['cats', 'dogs']
print(f"Total images: {len(full_ds):,}")
n_train = int(0.9 * len(full_ds))
train_ds, val_ds = random_split(full_ds, [n_train, len(full_ds) - n_train])
val_ds.dataset.transform = val_tfm # use val transform for val set
train_loader = DataLoader(train_ds, BATCH_SIZE, shuffle=True, num_workers=4, pin_memory=True)
val_loader = DataLoader(val_ds, BATCH_SIZE * 2, shuffle=False, num_workers=4, pin_memory=True)
# Model: EfficientNet-B0 (5.3M params, fast, accurate)
model = models.efficientnet_b0(weights=models.EfficientNet_B0_Weights.IMAGENET1K_V1)
model.classifier = nn.Sequential(
nn.Dropout(0.2),
nn.Linear(1280, 1), # binary classification → single logit
)
model = model.to(device)
# Training setup
optimizer = optim.AdamW([
{"params": model.features.parameters(), "lr": 1e-4},
{"params": model.classifier.parameters(), "lr": 1e-3},
], weight_decay=1e-4)
scheduler = optim.lr_scheduler.OneCycleLR(
optimizer, max_lr=[1e-4, 1e-3],
steps_per_epoch=len(train_loader), epochs=N_EPOCHS
)
loss_fn = nn.BCEWithLogitsLoss()
scaler = torch.cuda.amp.GradScaler()
best_acc = 0.0
for epoch in range(N_EPOCHS):
# Train
model.train()
for imgs, labels in train_loader:
imgs = imgs.to(device)
labels = labels.float().to(device).unsqueeze(1)
with torch.cuda.amp.autocast():
loss = loss_fn(model(imgs), labels)
scaler.scale(loss).backward()
scaler.step(optimizer)
scaler.update()
optimizer.zero_grad()
scheduler.step()
# Validate
model.eval()
correct = total = 0
with torch.no_grad():
for imgs, labels in val_loader:
imgs, labels = imgs.to(device), labels.to(device)
preds = (model(imgs).squeeze(1).sigmoid() > 0.5).long()
correct += (preds == labels).sum().item()
total += labels.size(0)
acc = correct / total
best_acc = max(best_acc, acc)
print(f"Epoch {epoch+1}/{N_EPOCHS} | Val Acc: {acc:.2%}")
if acc == best_acc:
torch.save(model.state_dict(), "best_dogs_cats.pt")
print(f"\nBest Val Accuracy: {best_acc:.2%}") # should be 99%+Tip
Tip
Practice Mini Project Dog vs Cat Classifier with Transfer Learning in small, isolated examples before integrating into larger projects. Breaking concepts into small experiments builds genuine understanding faster than reading alone.
Technical diagram.
Practice Task
Note
Practice Task — (1) Write a working example of Mini Project Dog vs Cat Classifier with Transfer Learning from scratch without looking at notes. (2) Modify it to handle an edge case (empty input, null value, or error state). (3) Share your solution in the Priygop community for feedback.
Quick Quiz
Common Mistake
Warning
A common mistake with Mini Project Dog vs Cat Classifier with Transfer Learning is skipping edge case testing — empty inputs, null values, and unexpected data types. Always validate boundary conditions to write robust, production-ready ai code.