A modular, training-free pipeline of off-the-shelf vision-language models plus object detection and self-verification modestly improves MCQ question answering from dense video captions.
Mm- au:towards multimodal understanding of advertisement videos
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Toward Scalable Video Narration: A Training-free Approach Using Multimodal Large Language Models
A modular, training-free pipeline of off-the-shelf vision-language models plus object detection and self-verification modestly improves MCQ question answering from dense video captions.