Summary of Meerkat: Audio-visual Large Language Model For Grounding in Space and Time, by Sanjoy Chowdhury et al.
Meerkat: Audio-Visual Large Language Model for Grounding in Space and Time
by Sanjoy Chowdhury, Sayan Nag, Subhrajyoti Dasgupta, Jun Chen, Mohamed Elhoseiny, Ruohan Gao, Dinesh Manocha
First submitted to arxiv on: 1 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
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 introduces Meerkat, a multi-modal large language model (LLM) that excels in understanding image and audio semantics spatially and temporally. Unlike previous LLMs focused on coarse-grained understanding, Meerkat leverages optimal transport-based modality alignment and cross-attention for consistency. The model is benchmarked on five challenging audio-visual tasks, including audio-referred image grounding, image-guided audio temporal localization, and audio-visual fact-checking. To evaluate Meerkat’s capabilities, the authors curate a large dataset AVFIT comprising 3M instruction tuning samples from open-source datasets. Meerkat achieves state-of-the-art performance on all tasks with up to 37.12% relative improvement. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new kind of computer model that can understand both images and sounds, and do things like identify objects in pictures based on their sounds. The model is called Meerkat, and it’s much better at this than previous models. It uses special techniques to make sure the image and sound parts work together correctly. Scientists tested Meerkat by giving it a series of tasks, such as identifying objects in pictures or pinpointing where specific sounds come from. Meerkat did very well on all these tasks, and this could be important for things like helping people with disabilities or creating better virtual assistants. |
Keywords
» Artificial intelligence » Alignment » Cross attention » Grounding » Instruction tuning » Large language model » Multi modal » Semantics
