Conventional deep semi-supervised learning methods, such as recursive clustering and training process, suﬀer from cumulative error and high computational complexity when collaborating with Convolutional Neural Networks. To this end, we design a simple but eﬀective learning mechanism that merely substitutes the last fully-connected layer with the proposed Transductive Centroid Projection (TCP) module. It is inspired by the observation of the weights in the ﬁnal classiﬁcation layer (called anchors) converge to the central direction of each class in hyperspace. Speciﬁcally, we design the TCP module by dynamically adding an ad hoc anchor for each cluster in one mini-batch. It essentially reduces the probability of the inter-class conﬂict and enables the unlabelled data functioning as labelled data. We inspect its eﬀectiveness with elaborate ablation study on seven public face/person classiﬁcation benchmarks. Without any bells and whistles, TCP can achieve signiﬁcant performance gains over most state-of-the-art methods in both fully-supervised and semi-supervised manners.