Leveraging over 30,000 images each with up to 89 labels collected by Recology---an integrated resource recovery company with both residential and commercial trash, recycling and composting services---the authors develop ContamiNet, a convolutional neural network, to identify contaminating material in residential recycling and compost bins. When training the model on a subset of labels that meet a minimum frequency threshold, ContamiNet preforms almost as well human experts in detecting contamination (0.86 versus 0.88 AUC). Recology is actively piloting ContamiNet in their daily municipal solid waste (MSW) collection to identify contaminants in recycling and compost bins to subsequently inform and educate customers about best sorting practices.