Keyframes First, Motion Second: Revisiting Compressed-Domain Video Captioning under Limited Supervision
cs.CV
Submitted: 2026-10-10
Updated: 2026-10-10
Comments: 13 pages including a 3-page supplementary appendix
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- ClipCap: CLIP Prefix for Image Captioning
- Modernising the Compressed-Domain Video Captioner: A Controlled Study of SigLIP2 and GPT-2 Substitutions
- CoPE-VideoLM: Leveraging Codec Primitives For Efficient Video Language Modeling
- DyCoke: Dynamic Compression of Tokens for Fast Video Large Language Models
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- VideoCoCa: Video-Text Modeling with Zero-Shot Transfer from Contrastive Captioners
- CodecSight: Leveraging Video Codec Signals for Efficient Streaming VLM Inference
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs