A Parallel Adaboost-Backpropagation Neural Network for Massive Image Dataset Classification

摘要: Image classification uses computers to simulate human understanding and cognition of images by automatically categorizing images. This study proposes a faster image approach that parallelizes the traditional Adaboost-Backpropagation (BP) neural network using MapReduce parallel programming model. First, we construct strong classifier assembling outputs 15 BP networks (which are individually regarded as weak classifiers) based on Adaboost algorithm. Second, design Map Reduce tasks for both Adaboost-BP feature extraction Finally, establish an automated model building Hadoop cluster. We use Pascal VOC2007 Caltech256 datasets train test The results superior those obtained or BP approaches. Our increased average accuracy rate approximately 14.5% 26.0% compared network, respectively. Furthermore, proposed requires less computation time scales very well evaluated speedup, sizeup scaleup. may provide foundation large-scale demonstrates practical value.