… about the dangers of using computer vision for describing people. In what follows, we consider
how these tensions are instantiated in the alt text practices and policies developed by two …

… computer vision research for object/theme identification, but to present a useful, fast, free
alt-text … The most similar system to our knowledge is alt-text bot [8], which also employed AI …

… ’s automatic alt text tool. When a blind or low-vision user encounters an image on Facebook,
the platform’s automatic alt text tool (AAT) uses a computer vision algorithm to process it …

… computer access to people who are BVI (Kloots 2016); however, while no official numbers
on alt text compliance for Twitter are yet available, alt text … that combine computer vision and …

… Figure 1: Examples of images and image descriptions from the Conceptual Captions dataset;
we start from existing alt-text descriptions, and automatically process them into Conceptual …

… is a complex and challenging problem that draws upon the fields of computer vision and
natural language processing. In an effort to be more accessible, several commercial platforms …

… Facebook’s Automated Alternative Text system uses a more conservative approach to
harnessing the output of computer vision by generating a series of tags that apply to an image with …

… To arrive at CC12M, we keep the image-text filtering intact, and relax the unimodal filters …
Second, in text-based filtering, we allow text between 3 and 256 words in the alt-text. We still …

… text recognition architectures and adapt to diverse computational constraints. In particular, we
explore a range of vision and vision… an alternative endto-end approach called text spotting […

… will require different ALT text according to the reason for, and location of, its use. ALT text
should be … ALT text should also be kept as accurate and succinct possible. Unnecessarily long …