To fully explore the potential of the weak labels, we propose to impose separate treatments of strong and weak annotations via a strong-weak dual-branch network, which discriminates the massive inaccurate weak supervisions from those strong ones.
This paper pays close attention to the cross-modality visible-infrared person re-identification (VI Re-ID) task, which aims to match human samples between visible and infrared modes.
(i) To learn more robust and discriminative features, we desire to adaptively fuse the contextual features from different layers.
Rule-based dialogue management is still the most popular solution for industrial task-oriented dialogue systems for their interpretablility.
However, the relation between the target extraction task and the target classification task has not been well exploited.
Deep part-based methods in recent literature have revealed the great potential of learning local part-level representation for pedestrian image in the task of person re-identification.
no code implementations • 3 Jan 2019 • Kyle B. Westfall, Michele Cappellari, Matthew A. Bershady, Kevin Bundy, Francesco Belfiore, Xihan Ji, David R. Law, Adam Schaefer, Shravan Shetty, Christy A. Tremonti, Renbin Yan, Brett H. Andrews, Joel R. Brownstein, Brian Cherinka, Lodovico Coccato, Niv Drory, Claudia Maraston, Taniya Parikh, José R. Sánchez-Gallego, Daniel Thomas, Anne-Marie Weijmans, Jorge Barrera-Ballesteros, Cheng Du, Daniel Goddard, Niu Li, Karen Masters, Héctor Javier Ibarra Medel, Sebastián F. Sánchez, Meng Yang, Zheng Zheng, Shuang Zhou
We summarize assessments of the precision and accuracy of our measurements as a function of signal-to-noise and other salient metrics, noting that additional analysis and discussion of emission-line diagnostics are provided in our companion paper, Belfiore et al.
Astrophysics of Galaxies
Second, to alleviate boundary artifacts of warped clothes and make the results more realistic, we employ a Try-On Module that learns a composition mask to integrate the warped clothes and the rendered image to ensure smoothness.
Furthermore, we show that the Euclidean norm appearing in the proximity function of the non-linear split feasibility problem can be replaced by arbitrary Bregman divergences.
Cross-entropy loss together with softmax is arguably one of the most common used supervision components in convolutional neural networks (CNNs).
We propose a robust elastic net (REN) model for high-dimensional sparse regression and give its performance guarantees (both the statistical error bound and the optimization bound).
Sparse coding with dictionary learning (DL) has shown excellent classification performance.
Practical face recognition has been studied in the past decades, but still remains an open challenge.
We propose the structured occlusion coding (SOC) to address occlusion problems.
Each dictionary atom is jointly learned with a latent vector, which associates this atom to the representation of different classes.
It is widely believed that the l1- norm sparsity constraint on coding coefficients plays a key role in the success of SRC, while its use of all training samples to collaboratively represent the query sample is rather ignored.
Recently the sparse representation based classification (SRC) has been proposed for robust face recognition (FR).