This paper investigates the impact of increasing the depth of convolutional networks (ConvNets) on their performance in large-scale image recognition tasks. The authors, Karen Simonyan and Andrew Zisserman from the Visual Geometry Group at the University of Oxford, focus on networks with very small (3×3) convolution filters, demonstrating that pushing the depth to 16–19 weight layers significantly improves accuracy compared to prior-art configurations. Their approach was successful in the ImageNet Challenge 2014, where they secured first and second places in the localization and classification tracks, respectively. The paper also shows that their models generalize well to other datasets, achieving state-of-the-art results. The authors have made their best-performing ConvNet models publicly available to facilitate further research on deep visual representations in computer vision. The paper is structured into sections covering the architecture design, training and evaluation procedures, experimental results, and comparisons with state-of-the-art methods.This paper investigates the impact of increasing the depth of convolutional networks (ConvNets) on their performance in large-scale image recognition tasks. The authors, Karen Simonyan and Andrew Zisserman from the Visual Geometry Group at the University of Oxford, focus on networks with very small (3×3) convolution filters, demonstrating that pushing the depth to 16–19 weight layers significantly improves accuracy compared to prior-art configurations. Their approach was successful in the ImageNet Challenge 2014, where they secured first and second places in the localization and classification tracks, respectively. The paper also shows that their models generalize well to other datasets, achieving state-of-the-art results. The authors have made their best-performing ConvNet models publicly available to facilitate further research on deep visual representations in computer vision. The paper is structured into sections covering the architecture design, training and evaluation procedures, experimental results, and comparisons with state-of-the-art methods.