Intrinsic Consistency Preservation With Adaptively Reliable Samples for Source-Free Domain Adaptation.

Unsupervised domain adaptation (UDA) aims to alleviate the domain shift by transferring knowledge learned from a labeled source dataset to an unlabeled target domain. Although UDA has seen promising progress recently, it requires access to data from both domains, making it problematic in source data-absent scenarios. In this article, we investigate a practical task source-free domain adaptation (SFDA) that alleviates the limitations of the widely studied UDA in simultaneously acquiring source and target data. In addition, we further study the imbalanced SFDA (ISFDA) problem, which addresses the intra-domain class imbalance and inter-domain label shift in SFDA. We observe two key issues in SFDA that: 1) target data form clusters in the representation space regardless of whether the target data points are aligned with the source classifier and 2) target samples with higher classification confidence are more reliable and have less variation in their classification confidence during adaptation. Motivated by these observations, we propose a unified method, named intrinsic consistency preservation with adaptively reliable samples (ICPR), to jointly cope with SFDA and ISFDA. Specifically, ICPR first encourages the intrinsic consistency in the predictions of neighbors for unlabeled samples with weak augmentation (standard flip-and-shift), regardless of their reliability. ICPR then generates strongly augmented views specifically for adaptively selected reliable samples and is trained to fix the intrinsic consistency between weakly and strongly augmented views of the same image concerning predictions of neighbors and their own. Additionally, we propose to use a prototype-like classifier to avoid the classification confusion caused by severe intra-domain class imbalance and inter-domain label shift. We demonstrate the effectiveness and general applicability of ICPR on six benchmarks of both SFDA and ISFDA tasks. The reproducible code of our proposed ICPR method is available at https://github.com/CFM-MSG/Code_ICPR.

Deep learning in multimedia healthcare applications: a review.

The increase in chronic diseases has affected the countries' health system and economy. With the recent COVID-19 virus, humanity has experienced a great challenge, which has led to make efforts to detect it and prevent its spread. Hence, it is necessary to develop new solutions that are based on technology and low cost, to satisfy the citizens' needs. Deep learning techniques is a technological solution that has been used in healthcare lately. Nowadays, with the increase in chips processing capabilities, increase size of data, and the progress in deep learning research, healthcare applications have been proposed to provide citizens' health needs. In addition, a big amount of data is generated every day. Development in Internet of Things, gadgets, and phones has allowed the access to multimedia data. Data such as images, video, audio and text are used as input of applications based on deep learning methods to support healthcare system to diagnose, predict, or treat patients. This review pretends to give an overview of proposed healthcare solutions based on deep learning techniques using multimedia data. We show the use of deep learning in healthcare, explain the different types of multimedia data, show some relevant deep learning multimedia applications in healthcare, and highlight some challenges in this research area.

NeedFull - a Tweet Analysis Platform to Study Human Needs During the COVID-19 Pandemic in New York State.

Governments and municipalities need to understand their citizens' psychological needs in critical times and dangerous situations. COVID-19 brings lots of challenges to deal with. We propose NeedFull, an interactive and scalable tweet analysis platform, to help governments and municipalities to understand residents' real psychological needs during those periods. The platform mainly consists of four parts: data collection module, data storage module, data analysis module and data visualization module. The four parts interact with each other and provide users with a thorough human needs analysis based on their queries. We employed the proposed platform to investigate the reaction of people in New York State to the ongoing worldwide COVID-19 pandemic.