Computer Vision for Data Scientists

Introduction

• Computer vision introduction
• What you should know

1. Computer Vision

• What is computer vision?
• A history of computer vision
• Limitations of traditional CV techniques
• ImageNet
• The deep learning revolution

2. Introduction to Convolutional Neural Networks

• Overview of CNNs
• Why CNNs?
• Convolutional layers
• Types of convolutions
• Pooling layers
• Activation functions
• Fully connected layers

3. How Networks Are Trained

• Supervised learning and loss functions
• Backpropagation in CNNs
• Optimization techniques
• Regularization and data augmentation

4. The Evolution of CNN Architectures

• LeNet
• AlexNet
• VGG
• ResNet
• MobileNetV1
• MobileNetV2
• MobileNetV3
• EfficientNet

5. Transfer Learning

• Introduction to transfer learning
• Types of transfer learning
• Steps in feature extracting and fine-tuning
• Best practices for transfer learning

6. PyTorch Crash Course

• Setting up the environment
• Dataset and DataLoader
• Training setup
• The training loop
• Testing and evaluation
• Inference

7. Hands-on Transfer Learning with SuperGradients

• Introduction to SuperGradients
• The trainer
• Required training params
• Optional training params
• Training the model
• Predicting with the model
• How to solve almost any image classification problem with SG

8. Training Tricks

• Exponential moving average
• Weight averaging
• Batch accumulation
• Precise BatchNorm
• Zero weight decay on BatchNorm and bias
• Training tricks in SuperGradients

Conclusion

• Next steps