← All works
Research paper

Deep Visual-Semantic Alignments for Generating Image Descriptions

At a glance

Connected words with image regions to generate descriptions of scenes and their details.

In one paragraph

Developed a neural network that connects words with regions of images and generates descriptions, helping computers describe both entire scenes and specific details within them.

Authors 2

  1. Andrej Karpathy
  2. Fei-Fei Li