Additional Unlabeled Samples Research Articles

Self‐training with one‐shot stepwise learning method for person re‐identification

SummaryPerson re‐identification (Re‐ID) aims at identifying the same person across multiple non‐overlapping camera views. A number of existing methods have been presented for this task in a fully‐supervised manner that requires a large amount of training annotations. However, obtaining high quality labels is extremely time consuming and expensive. In this article, we focus on the semi‐supervised person Re‐ID and propose a one‐shot stepwise learning method to address the above issue. It exploits only one labeled data along with additional unlabeled samples to gradually but steadily improving the discriminative capability of the feature representation. Specifically, we first construct labeled data portion to train Re‐ID model. Then we fine‐tune the overall system by the following two steps iteratively: (1) assigning the estimated labels to the unlabeled portion; (2) updating the network parameters according to the selected data. During the propagation process, different from conventional sampling method, we propose a novel dynamic sampling strategy to enlarge the pseudo‐labeled subset step by step to make the pseudo labels more reliable. On Market‐1501, DukeMTMC‐ReID and MARS datasets, we conducted extensively experiments to demonstrate that our proposed method contributes indispensably and achieves a very competitive Re‐ID performance.

Semi-supervised person re-identification using multi-view clustering

Person Re-Identification (Re-Id) is a challenging task focusing on identifying the same person among disjoint camera views. A number of deep learning algorithms have been reported for this task in fully-supervised fashion which requires a large amount of labeled training data, while obtaining high quality labels for Re-Id is extremely time consuming. To address this problem, we propose a semi-supervised Re-Id framework by using only a small portion of labeled data and some additional unlabeled samples. This paper approaches the problem by constructing a set of heterogeneous Convolutional Neural Networks (CNNs) fine-tuned using the labeled portion, and then propagating the labels to the unlabeled portion for further fine-tuning the overall system. In this work, label estimation is a key component during the propagation process. We propose a novel multi-view clustering method, which integrates features of multiple heterogeneous CNNs to cluster and generate pseudo labels for unlabeled samples. Then we fine-tune each of the multiple heterogeneous CNNs by minimizing an identification loss and a verification loss simultaneously, using training data with both true labels and pseudo labels. The procedure is iterated until the estimation of pseudo labels no longer changes. Extensive experiments on three large-scale person Re-Id datasets demonstrate the effectiveness of the proposed method.

Global Rademacher Complexity Bounds: From Slow to Fast Convergence Rates

Previous works in literature showed that performance estimations of learning procedures can be characterized by a convergence rate ranging between $$O(n^{-1})$$O(n-1) (fast rate) and $$O(n^{-{1}/{2}})$$O(n-1/2) (slow rate). In order to derive such result, some assumptions on the problem are required; moreover, even when convergence is fast, the constants characterizing the bounds are often quite loose. In this work, we prove new Rademacher complexity (RC) based bounds, which do not require any additional assumptions for achieving a fast convergence rate $$O(n^{-1})$$O(n-1) in the optimistic case and a slow rate $$O(n^{-{1}/{2}})$$O(n-1/2) in the general case. At the same time, they are characterized by smaller constants with respect to other state-of-the-art RC fast converging alternatives in literature. The results proposed in this work are obtained by exploiting the fundamental work of Talagrand on concentration inequalities for product measures and empirical processes. As a further issue, we also provide the extension of the results to the semi-supervised learning framework, showing how additional unlabeled samples allow improving the tightness of the derived bound.

The effect of unlabeled samples in reducing the small sample size problem and mitigating the Hughes phenomenon

The authors study the use of unlabeled samples in reducing the problem of small training sample size that can severely affect the recognition rate of classifiers when the dimensionality of the multispectral data is high. The authors show that by using additional unlabeled samples that are available at no extra cost, the performance may be improved, and therefore the Hughes phenomenon can be mitigated. Furthermore, by experiments, they show that by using additional unlabeled samples more representative estimates can be obtained. They also propose a semiparametric method for incorporating the training (i.e., labeled) and unlabeled samples simultaneously into the parameter estimation process. >