Fig. 4From: Learning deep networks with crowdsourcing for relevance evaluationFine tuning. The reconstruction of VGG-16 to make an autoencoder for fine-tuningBack to article page