# LightAVSeg: Lightweight Audio-Visual Segmentation

arXiv: https://arxiv.org/abs/2605.08805 · HTML: https://mlsanity.com/paper/2605.08805/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 34 references that are in the map.
- First posted on arXiv 9 May 2026.

## Builds on (most cited first, 26 of 26)

- [Fully Convolutional Networks for Semantic Segmentation](https://mlsanity.com/paper/1411.4038/) (2014)
- [DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs](https://mlsanity.com/paper/1606.00915/) (2016)
- [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://mlsanity.com/paper/2105.15203/) (2021)
- [Identity Mappings in Deep Residual Networks](https://mlsanity.com/paper/1603.05027/) (2016)
- [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://mlsanity.com/paper/2102.12122/) (2021)
- [PVT v2: Improved Baselines with Pyramid Vision Transformer](https://mlsanity.com/paper/2106.13797/) (2021)
- [Holistically-Nested Edge Detection](https://mlsanity.com/paper/1504.06375/) (2015)
- [PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition](https://mlsanity.com/paper/1912.10211/) (2019)
- [Look, Listen and Learn](https://mlsanity.com/paper/1705.08168/) (2017)
- [Deeply-Supervised Nets](https://mlsanity.com/paper/1409.5185/) (2014)
- [Audio-Visual Scene Analysis with Self-Supervised Multisensory Features](https://mlsanity.com/paper/1804.03641/) (2018)
- [Objects that Sound](https://mlsanity.com/paper/1712.06651/) (2017)
- [Learning to Localize Sound Source in Visual Scenes](https://mlsanity.com/paper/1803.03849/) (2018)
- [Localizing Visual Sounds the Hard Way](https://mlsanity.com/paper/2104.02691/) (2021)
- [Audio-Visual Segmentation](https://mlsanity.com/paper/2207.05042/) (2022)
- [Fast-SCNN: Fast Semantic Segmentation Network](https://mlsanity.com/paper/1902.04502/) (2019)
- [TopFormer: Token Pyramid Transformer for Mobile Semantic Segmentation](https://mlsanity.com/paper/2204.05525/) (2022)
- [Audio-Visual Segmentation with Semantics](https://mlsanity.com/paper/2301.13190/) (2023)
- [AVSegFormer: Audio-Visual Segmentation with Transformer](https://mlsanity.com/paper/2307.01146/) (2023)
- [Look, Listen, and Attend: Co-Attention Network for Self-Supervised Audio-Visual Representation Learning](https://mlsanity.com/paper/2008.05789/) (2020)
- [CTVIS: Consistent Training for Online Video Instance Segmentation](https://mlsanity.com/paper/2307.12616/) (2023)
- [A Comedy of Estimators: On KL Regularization in RL Training of LLMs](https://mlsanity.com/paper/2512.21852/) (2025)
- [AVS-Mamba: Exploring Temporal and Multi-modal Mamba for Audio-Visual Segmentation](https://mlsanity.com/paper/2501.07810/) (2025)
- [Sound event detection via dilated convolutional recurrent neural networks](https://mlsanity.com/paper/1911.10888/) (2019)
- [MobileInst: Video Instance Segmentation on the Mobile](https://mlsanity.com/paper/2303.17594/) (2023)
- [Complementary and Contrastive Learning for Audio-Visual Segmentation](https://mlsanity.com/paper/2510.10051/) (2025)

## Abstract (arXiv)

Audio-Visual Segmentation (AVS) targets pixel level localization of sounding emitting objects in videos. However, existing models rely on dense cross-modal attention with quadratic computational cost, limiting their suitability for resource efficient deployment. Most efficiency oriented methods focus on backbone reduction and overlook the interaction module as the primary bottleneck. This paper proposes LightAVSeg, a lightweight framework that replaces heavy attention with a decoupled design for semantic filtering and spatial grounding, resulting in interaction costs that scale linearly with spatial resolution. Furthermore, we introduce an auxiliary alignment loss to enforce semantic consistency during training with zero inference overhead. Extensive experiments demonstrate that LightAVSeg achieves a new state-of-the-art among lightweight methods: with 20.5M parameters ~1/7 of AVSegFormer), it reaches 50.4 mIoU on the MS3 benchmark and enables efficient inference on a mobile processor.