Preface
Lightweight network design is a deep learning model design method for resource-constrained environments such as mobile devices. Here are some common lightweight network design methods:
- Network pruning: Remove redundant connections and parameters in the neural network to achieve model compression and acceleration.
- Group convolution: Decompose the convolution operation into several smaller convolution operations and apply them to different channels of the input, thereby reducing the amount of calculation.
- Depthwise separable convolution: Decompose the standard convolution into two steps: depth convolution and point-wise convolution, which can greatly reduce the amount of calculation in most cases.
- Cross-layer connection: Increase the depth and complexity of the neural network through connections across multiple levels, while reducing the number of parameters that need to be trained.
- Modular design: Decompose the neural network into multiple reusable modules to improve the adjustability and adaptability of the model.
TraditionalYOLOv8 series, Backbone adopts It is a relatively complex C2f network structure, which greatly increases the calculation amount of the model, the detection speed is slow, and the application is limited. In some real situations In application scenarios such as mobile or embedded devices, such large and complex models are difficult to apply. In order to solve this problem, this chapter adopts ShuffleNetV2</