Image recognition is a general term often used to describe the task of identifying the content in an image. It may encompass different problems: usually image classification (identifying the main object or scene in an image) is what is meant by image recognition. For instance, recognizing handwritten digits, or recognizing faces, or identifying that an image contains a car. It sometimes includes detection and other tasks, but typically recognition implies assigning a label to the image or to objects within the image. Modern image recognition is achieved with deep learning, especially CNNs trained on large datasets. Historically, image recognition was approached with hand-engineered features and classifiers (like SIFT + SVMs for object recognition), but now end-to-end learning predominates. Performance in image recognition has dramatically improved, enabling real-world applications like photo tagging, visual search, and aiding autonomous driving perception.