🧪 Test?View on arXiv
StrAD: A Streaming Method and Benchmark for Audio Description Generation for Long-form Videos
Not specified in the content
streamingaudio descriptionvideo captioningaccessibility
2608.12549
Builder Relevance
6h ago80%
Abstract
The paper presents StrAD, a benchmark and method for generating audio descriptions for long-form videos, addressing accessibility for blind and low-vision individuals.
Reality Card
Core Claim
StrAD is the first streaming approach for generating audio descriptions on the fly without requiring ground-truth timestamps.
Method / Result
StrAD-FT sets the state of the art on CMD-AD with a CIDEr score of 36.3.
Limitations
Both StrAD-FT and StrAD-Zero exhibit limitations in temporal localization and narrative coherence.