TensorFlow 2. 0 Computer Vision Cookbook

Computer vision is the field that allows machines to interpret and process digital images and video. This cookbook takes a recipe-by-recipe approach, with each section tackling a separate practical task in TensorFlow. It opens with the fundamentals of deep learning for vision and introduces key TensorFlow 2.x tools, including the Keras API and the tf.data.Dataset API.

From there, the recipes cover image classification, transfer learning, image enhancement and style transfer, and object detection. Autoencoders are applied to image denoising and to building inverse image search indexes. Along the way, the book explains architectures such as convolutional neural networks, region-based CNNs, VGGNet, and YOLO, and it offers troubleshooting tips for common problems that come up while building vision applications.

Later chapters move into Generative Adversarial Networks, video analysis, and AutoML, and they close with methods for improving model performance. Readers learn to detect objects with YOLOv3, predict age and gender from photos, segment images, recognize emotions in video and live streams, reuse TensorFlow Hub models, and generate image captions with CNNs and RNNs.

The book is intended for computer vision developers, engineers, and deep learning practitioners who want practical solutions to recurring vision problems. Readers should already know Python and have a basic understanding of computer vision.