BEEMS: Boosting Machine Vision Efficiency via Computation Graph-Based Memory Smoothing
Hanjing Shen, Fangxin Liu, Jian Liu, Li Jiang, Haibing Guan
摘要
With the rapid advances of deep learning-based computer vision (CV) technology, digital images are increasingly processed not by humans, but by downstream CV algorithms. In particular, the growing popularity of vision foundation models has heightened interest in deploying these models on edge devices. However, limited memory remains a key bottleneck, making memory footprint reduction essential. Mainstream model customization methods often require intensive deployment efforts and can severely degrade accuracy. Moreover, existing deep learning frameworks generally do not prioritize memory optimization. Existing memory management schemes face practical limitations, including layer-wise memory imbalance, high management overhead, and volatile memory budgets.
问 Lune
几分钟看清这篇 Paper 的位置,不用再花一下午。
Lune 能读引用它和与它最相关的顶会 Paper,讲清它说了什么、后来怎样、哪里站不住,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
相关 Paper
- FlexNN: Efficient and Adaptive DNN Inference on Memory-Constrained Edge DevicesXiangyu Li, Yuanchun Li, Yuanzhe Li, Ting Cao 等MobiCom 2024 · 40 次引用
- Activation Map Compression through Tensor Decomposition for Deep LearningLe-Trung Nguyen, Aël Quélennec, Enzo Tartaglione, Samuel Tardieu 等NeurIPS 2024 · 7 次引用
- Q-CapsNets: A Specialized Framework for Quantizing Capsule NetworksAlberto Marchisio, Beatrice Bussolino, Alessio Colucci, Maurizio Martina 等DAC 2020 · 19 次引用
- ALF: Autoencoder-based Low-rank Filter-sharing for Efficient Convolutional Neural NetworksAlexander Frickenstein, Manoj Rohit Vemparala, Nael Fasfous, Laura Hauenschild 等DAC 2020 · 5 次引用
- Gemel: Model Merging for Memory-Efficient, Real-Time Video Analytics at the EdgeArthi Padmanabhan, Neil Agarwal, Anand P. Iyer, Ganesh Ananthanarayanan 等NSDI 2023 · 94 次引用
