What Is Image Captioning? Generating a Written Description of a Picture
The task of automatically writing an accurate, natural-language description of what's actually happening in an image.
Describing an image in words
Image captioning is the task of automatically generating a natural-language description of an image's content — not just listing detected objects, but describing what's actually happening and how the elements in the scene relate to each other.
Why this is harder than simple object detection
Identifying that an image contains "a dog" and "a ball" is straightforward object detection. Generating "a dog catching a ball mid-air in a park" requires understanding the relationship and action connecting those objects — genuine scene understanding, not just a list of detected items.
Where captioning gets used
Beyond generating standalone descriptions, image captioning underpins accessibility features like alt-text generation for visually impaired users, and it's a core capability inside broader vision-language models that need to describe or reason about images as part of a larger conversation.
Frequently Asked Questions
What is image captioning?
The task of automatically generating a natural-language description of an image's content, capturing not just the objects present but the action and relationships between them.
How is image captioning different from object detection?
Object detection identifies what's present in an image; captioning describes what's actually happening, requiring understanding of how the detected elements relate to and interact with each other.