8 Introduction to deep learning for computer vision

Computer vision is the earliest and biggest success story of deep learning. Every day, you’re interacting with deep vision models—via Google Photos, Google image search, YouTube, video filters in camera apps, OCR software, and many more. These models are also at the heart of cutting-edge research in autonomous driving, robotics, AI-assisted medical diagnosis, autonomous retail checkout systems, and even autonomous farming.

Computer vision is the problem domain that led to the initial rise of deep learning between 2011 and 2015. A type of deep learning model called convolutional neural networks started getting remarkably good results on image classification competitions around that time, first with Dan Ciresan winning two niche competitions (the ICDAR 2011 Chinese character recognition competition and the IJCNN 2011 German traffic signs recognition competition), and then more notably in fall 2012 with Hinton’s group winning the high-profile ImageNet large-scale visual recognition challenge. Many more promising results quickly started bubbling up in other computer vision tasks.

8.1 Introduction to convnets

8.1.1 The convolution operation

8.1.2 The max-pooling operation

8.2 Training a convnet from scratch on a small dataset

8.2.1 The relevance of deep learning for small-data problems

8.2.2 Downloading the data

8.2.3 Building the model

8.2.4 Data preprocessing

8.2.5 Using data augmentation

8.3 Leveraging a pretrained model

8.3.1 Feature extraction with a pretrained model

8.3.2 Fine-tuning a pretrained model

Summary