2017/05/25 by Bangalore Ravi Kiran, Kiran, B Ravi, Arindam Das +4
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Human Pose and Action Recognition #Image Enhancement Techniques #Machine Learning (stat.ML) #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1705.09339
openalex publication_date 2017/05/25 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Background-Foreground classification is a well-studied problem in computer\nvision. Due to the pixel-wise nature of modeling and processing in the\nalgorithm, it is usually difficult to satisfy real-time constraints. There is a\ntrade-off between the speed (because of model complexity) and accuracy.\nInspired by the rejection cascade of Viola-Jones classifier, we decompose the\nGaussian Mixture Model (GMM) into an adaptive cascade of Gaussians(CoG). We\nachieve a good improvement in speed without compromising the accuracy with\nrespect to the baseline GMM model. We demonstrate a speed-up factor of 4-5x and\n17 percent average improvement in accuracy over Wallflowers surveillance\ndatasets. The CoG is then demonstrated to over the latent space representation\nof images of a convolutional variational autoencoder(VAE). We provide initial\nresults over CDW-2014 dataset, which could speed up background subtraction for\ndeep architectures.\n