… The cost of system failure in our use case is much higher than in existing auto-captioning
systems – eg, blind users could be misled to make inappropriate comments about photos in …

… SMALLCAP, an image captioning model, prompted with captions retrieved from an external
datastore of text, based on the input image. This formulation of image captioning enables a …

… text pairs. To bridge the gap between the limited diversity of generated captions and the
high diversity of alternative text (alt-text… model pretrained on unlabeled text corpus. Using …

… Our goal here is to optimize the text generation module of the image captioning model. We
used the RNN architecture as our text generation module and our goal here is to optimize the …

… captioning competition. Our work provides a step towards improved assistive image captioning
… When alternative text is either missing or provides non-meaningful content, there is a …

… AI image captioning systems … as image labeling with objective ground truth answers,
and it is unclear how human input can be combined with automated image captioning for alt-text …

… Microsoft, which offers image captioning services to “improve accessibility” through Azure
AI, appears to include gender in some instances and not others [44]. As of April 2021, Amazon …

… attentive linear transformation (ALT) for automatic generation of image captions. Instead of
… , ALT learns to attend to the high-dimensional transformation matrix from the image feature …

… ’s Automatic AltText system originally aimed to generate image tags that describe the prominent
objects in an image [… -language image description, along with providing tags grouped by …

… Meanwhile, we observe that captions generated by previous image captioning works [6], [14]
are logical and fluent in language, but they often produce some inaccurate details. …