Summary of Infp: Audio-driven Interactive Head Generation in Dyadic Conversations, by Yongming Zhu et al.
INFP: Audio-Driven Interactive Head Generation in Dyadic Conversations
by Yongming Zhu, Longhao Zhang, Zhengkun Rong, Tianshu Hu, Shuang Liang, Zhipeng Ge
First submitted to arxiv on: 5 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel audio-driven head generation framework called INFP, which enables seamless dyadic interaction between humans and socially intelligent agents. The model dynamically alternates between speaking and listening states, guided by input dyadic audio. Unlike previous works, INFP doesn’t require manual role assignment or explicit role switching. It consists of two stages: Motion-Based Head Imitation and Audio-Guided Motion Generation. The first stage learns to project facial communicative behaviors from real-life conversation videos into a low-dimensional motion latent space and animates a static image. The second stage maps input dyadic audio to motion latent codes through denoising, generating an audio-driven head in interactive scenarios. The authors also introduce DyConv, a large-scale dataset of rich dyadic conversations collected from the Internet. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine having a conversation with a smart robot that can understand and respond like a person. This paper helps make that possible by creating a new way for robots to generate faces that look like they’re listening or talking. Instead of just showing one face, this method can switch between speaking and listening states based on what’s being said. The robot uses real-life conversations to learn how to move its “face” in different ways. This helps the conversation feel more natural and smooth. To test this new way of generating faces, the authors created a big dataset of conversations that they used to train their model. |
Keywords
» Artificial intelligence » Latent space
