Previous lesson. Capstone Blueprints (public portfolio practice) · Next lesson. Project 2: Sentiment Analysis on Reviews
Project 1: Image Classification (CIFAR-10)
Build a CNN to classify images into 10 categories with high accuracy.
Difficulty
Advanced
Time Estimate
1-2 weeks
Skills You'll Practice
- CNNs
- Transfer Learning
- Data Augmentation
- Model Optimization
Learning Objectives
- Build advanced CNN architectures
- Apply transfer learning with pre-trained models
- Use data augmentation effectively
- Optimize models for performance
- Handle complex image data
Dataset
CIFAR-10
- Built into Keras:
keras.datasets.cifar10 - 50,000 training images, 10,000 test images
- 32x32 color images in 10 classes
Project Steps
- Load and explore CIFAR-10 dataset
- Build baseline CNN
- Apply data augmentation
- Use transfer learning (ResNet, VGG, etc.)
- Optimize hyperparameters
- Achieve >90% accuracy
- Compare different architectures
Expected Deliverables
- Complete notebook with analysis
- Trained model achieving >90% accuracy
- Comparison of different architectures
- Deployment-ready model
Resources
Previous lesson. Capstone Blueprints (public portfolio practice) · Next lesson. Project 2: Sentiment Analysis on Reviews