Selective Depthwise Separable Convolution for Lightweight Joint Source-Channel Coding in Wireless Image Transmission
The research proposes a configurable framework called DSC-JSCC to reduce computational complexity in deep learning-based joint source-channel coding for wireless image transmission. It systematically investigates replacing standard convolutional layers with depthwise separable convolutions at different ratios and positions. The findings show that replacing intermediate layers offers a good balance between model performance and reduced computational cost, making the resulting models suitable for resource-constrained edge devices.