Open-Source Code for Overlay Detection in Website Screenshots using Deep Learning
1 min read
We provide the open-source Python code used to train and evaluate deep learning models for detecting overlays in website screenshots, as presented in our ICAR'15 paper. The code, implemented using Keras and TensorFlow, includes the implementation and training procedures for three Convolutional Neural Network (CNN) architectures: a Simple CNN baseline, MobileNetV2, and ResNet50 (both using transfer learning). This repository allows researchers to replicate our experiments, test the models on the provided dataset or their own data, and potentially integrate the overlay detection logic into web scraping or analysis tools. Access the code via the link below.
git clone git@github.com:goker-dev/overlay.git