VGG-16 is a deep convolutional neural network architecture consisting of 16 weight layers, originally developed for large-scale image recognition, characterized by its use of small 3×3 convolutional filters stacked in increasing depth. In both crowd counting and WiFi CSI-based human sensing, VGG-16 serves as a powerful feature extractor and backbone network, leveraging its deep hierarchical representations to capture complex spatial patterns in images or signal data. Its significance lies in its strong transfer learning capability, as models pre-trained on large datasets such as ImageNet can be fine-tuned for domain-specific tasks, and it is commonly employed alongside variants such as VGG-19, which extends the architecture to 19 layers for potentially richer feature extraction.
Source Papers
- A survey of recent advances in CNN-based single image crowd counting and density estimation ↗ — A survey of recent advances in CNN-based single image crowd
- Recent trends in crowd analysis: A review ↗ — Recent trends in crowd analysis: A review
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered