ImageNet is a dataset of images annotated mainly with labels for object categories, such as 'dog', 'cat', etc. The labeling follows the so called WordNet hierarchy, which takes into account the semantic relationships between categories (for example, the category 'poodle' is nested within the category 'dog'). This dataset is widely used in Computer Vision to develop image classification models. A subset of ImageNet has been annotated with bounding boxes around objects, and can therefore be used to develop object detection models. This subset, known as ImageNet-1k, spans 1000 object classes, contains over 1.2 million training images, 50 thousand validation images, and 100 thousand test images.