AI Insight
Researchers have developed a new method called Channel Group-Shared (CGS) low-rank approximation to reduce the storage requirements of large-kernel convolutional neural networks (CNNs) for image recognition on smartphones and other edge devices. The technique addresses a critical bottleneck where pointwise convolutions consume over 87% of model parameters, making deployment on memory-constrained devices impractical. By using a structured parameter-sharing strategy based on Singular Value Decomposition, the method significantly compresses these networks while maintaining competitive performance, enabling high-quality AI vision models to run efficiently on devices with limited RAM.
Why it matters
This advancement could enable sophisticated AI image recognition capabilities on everyday smartphones and IoT devices without requiring cloud connectivity or expensive hardware. By reducing storage costs and memory demands, the technology makes it feasible to deploy powerful computer vision models directly on consumer electronics, potentially improving privacy, reducing latency, and enabling AI functionality in areas with limited internet connectivity.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Large-kernel Convolutional Neural Networks (CNNs) deliver remarkable performance in vision tasks by significantly expanding receptive fields, yet their quadratic parameter growth critically impedes storage-efficient edge deployment. While existing efficient architectures adopt parameter-efficient depthwise separable convolution backbones that leverage techniques like low-rank approximation and weight sharing to compress depthwise convolutions, we identify a critical oversight: pointwise convolutions dominate parameter volume (>87% in models like RepLKNet-31B) and constitute the primary deployment bottleneck on resource-constrained edge devices. This results in prohibitive storage costs and severe memory-loading constraints on resource-limited devices (e.g., smartphones with 4-12 GB Random Access Memory (RAM)). To overcome this, we propose Channel Group-Shared (CGS) low-rank approximation, a novel Singular Value Decomposition (SVD)-based parameter-sharing strategy. CGS constructs a structured low-rank paradigm isomorphic to SVD decomposition, comprising shared (high-parameter-cost) down/up-projection matrices across channel groups within a layer and channel-group-specific (low-parameter-cost) scalable diagonal matrices. This group-sharing design achieves significant parameter reduction. Extensive experiments demonstrate that large-kernel CNNs (RepLKNet, ConvNeXt, SLaK) enhanced with CGS strike an empirically favorable balance between competitive performance and substantially reduced storage costs. Crucially, by alleviating storage constraints, reducing memory bandwidth pressure during loading, and minimizing model loading latency, CGS enables the feasible deployment of pre-trained large-kernel CNN models on edge devices, thereby bridging the gap between high-performance vision models and practical edge deployment.
Source: Group-Shared Low-Rank Approximation for Mobile-Efficient Pointwise Convolutions in Large-Kernel CNNs