LongVU adaptively compresses long video tokens using DINOv2-based frame deduplication, text-guided cross-modal selection, and temporal spatial reduction to improve video-language understanding in MLLMs with minimal detail loss.
Therefore, we decide not to include it in our default setting
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding
LongVU adaptively compresses long video tokens using DINOv2-based frame deduplication, text-guided cross-modal selection, and temporal spatial reduction to improve video-language understanding in MLLMs with minimal detail loss.