New Semantic Frame Interpolation task and benchmark introduced

Researchers have introduced Semantic Frame Interpolation (SFI), a new task for generating intermediate video content of varying lengths based on start and end frames and a text prompt. This work proposes the SemFi model, built upon Wan2.1 and incorporating a Mixture-of-LoRA module for content consistency across different frame lengths. To support SFI, the paper also introduces SFI-300K, the first general-purpose dataset and benchmark specifically designed for this task, along with new evaluation metrics. AI

RANK_REASON This is a research paper introducing a new task, model, and dataset in computer vision. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Hugging Face
• Mixture-of-LoRA
• Semantic Frame Interpolation
• SFI-300K
• Wan2.1
• Yijia Hong

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →