ImageNet

ImageNet is a large-scale visual database containing over 14 million labeled images organized into thousands of categories, widely used for training and benchmarking computer vision models. Many object detection and image classification systems used in accessibility applications—such as automatic alt text generators, audio description systems, and visual assistance apps for blind users—are trained on ImageNet or use models pre-trained on this dataset. The annual ImageNet Large Scale Visual Recognition Challenge (ILSVRC) drove major advances in deep learning that now power accessibility features.