EnCNN-UPMWS: Waste Classification by a CNN Ensemble Using the UPM Weighting Strategy

The accurate and effective classification of household solid waste (HSW) is an indispensable component in the current procedure disposal. In this paper, a novel ensemble learning model called EnCNN-UPMWS, which based on convolutional neural networks (CNNs) unequal precision measurement weighting strategy (UPMWS), proposed for HSW via images. First, three state-of-the-art CNNs, namely GoogLeNet, ResNet-50, MobileNetV2, are used as ingredient classifiers to separately predict obtain predicted probability vectors, significant elements that affect prediction performance by providing complementary information about patterns be classified. Then, UPMWS introduced determine weight coefficients models. actual one-hot encoding labels validation set vectors from CNN creatively calculate weights each classifier during training phase, can bring aggregated vector closer target label improve model. was applied two datasets, TrashNet (an open-access dataset) FourTrash, constructed collecting total 47,332 common images containing four types (wet waste, recyclables, harmful dry waste). experimental results demonstrate effectiveness method terms its accuracy F1-scores. Moreover, it found simply effectively enhance model, has potential applications similar tasks learning.