Shulou(Shulou.com)11/24 Report--

According to news from CTOnews.com on November 14, the 31st International Multimedia Conference (ACM International Conference on Multimedia,ACMMM) was held in Ottawa, Canada, from October 29 to November 2. The Chinese team gained a lot from this meeting.

CTOnews.com Note: ACM International Multimedia Conference (ACMMM) has been held once a year since it was first held in 1993. It has become a top conference in the multimedia field, and it is also a class An international academic recommended by the Chinese computer Society.

This year's ACMMM conference received a total of 3072 submitted papers and accepted 902, with an enrollment rate of 29.3%. Among them, the team of Zhejiang University won the Best Paper Award (Best Paper Award), the team of Nanjing University of Aeronautics and Astronautics won the Honorary nomination Award (Honorable Mention Award), and the team of Tsinghua University won the Brave creativity Award (Brave New Idea Award).

The title of this best paper award is "CATR: Combinatorial-Dependence Audio-Queried Transformer for Audio-Visual Video Segmentation" (CATR: combination dependent Audio query Converter for Video Segmentation).

Paper link: https://arxiv.org/abs/2309.09709

Open source address: https://github.com/aspirinone/CATR.github.io

Author unit: Zhejiang University;Finvolution Group

This paper mainly introduces video segmentation guided by cross-modal information, and proposes a video segmentation framework based on combinatorial dependency and audio query (CATR).

In this scheme, the pixel-level mask marks the sound object in the video, and associates the sound and the sound-making object, so as to identify and segment the person who makes the sound in the video.

Honorary nomination Award RefineTAD: Learning Proposal-free Refinement for Temporal Action Detection

Paper link: https://dl.acm.org/doi/pdf/10.1145/3581783.3611872

Author unit: Nanjing University of Aeronautics and Astronautics;Nanjing University

This paper mainly studies time motion detection (TAD). Through the new location optimization framework, the location refinement process is decoupled from the traditional motion detection methods, and multi-scale location refinement information is generated at each time point.

Bravely Creative Award Semantics2Hands: Transferring Hand Motion Semantics between Avatars

Paper link: https://arxiv.org/abs/2308.05920

Open source address: https://github.com/abcyzj/Semantics2Hands

Author unit: Tsinghua University;Tsinghua University Beijing National Research Center for Information Science and Technology

This paper mainly carries on the hand movement transfer on the premise of keeping the hand action semantics of the virtual characters in the animation and human-computer interaction.

More information can be found at https://www.acmmm2023.org/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.