Search

Generative Adversarial Networks (GANs) can synthesize more and more realistic images. However, one fundamental issue hinders their practical applications: the incapability of encoding real samples in the latent space. Many semantic image editing applications rely on inverting the given image into the latent space and then manipulating inverted code. One possible solution is to learn an encoder for GAN via Variational Auto-Encoder. However, the entanglement of the latent space poses a major challenge for learning the encoder. To tackle the challenge and enable inference in GANs, we propose a novel method named Latently Invertible Autoencoder (LIA). In LIA, an invertible network and its inverse mapping are symmetrically embedded in the latent space of an autoencoder. The decoder of LIA is first trained as a standard GAN with the invertible network, and then the encoder is learned from a disentangled autoencoder by detaching the invertible network from LIA. It thus avoids the entanglement problem caused by the latent space. Extensive experiments on the FFHQ face dataset and three LSUN datasets validate the effectiveness of LIA for the image inversion and its applications. Code and models are available at https://github.com/genforce/lia. [ABSTRACT FROM AUTHOR]

Artesunate, a derivative of artemisinin, has been reported to have anti-inflammatory property. However, few studies showed the protective effects of artesunate on lung injury. In this study, we aimed to investigate the effects of artesunate on LPS-induced lung injury in mice. The mice were treated with artesunate 1 h before or after LPS treatment. The effects of artesunate on lung MPO activity and malondialdehyde (MDA) content were detected. The lung wet/dry radio and the numbers of inflammatory cells in BALF were also measured. ELISA was used to evaluate the levels of TNF-α, IL-1β, and IL-6 in BALF. Western blot analysis was adapted to detect TLR4 and Nrf2 signaling pathways. The results showed that artesunate protected against LPS-induced ALI by decreasing the numbers of inflammatory cells, lung edema, MPO activity, and MDA content. Furthermore, artesunate significantly inhibited the levels of TNF-α, IL-1β, and IL-6. Artesunate also inhibited LPS-induced IL-6 and IL-8 production in the A549 cells. In addition, artesunate dose-dependently suppressed LPS-induced TLR4 expression and NF-κB activation. The expression of Nrf2 and HO-1 were also up-regulated by artesunate. The data suggest that artesunate possesses anti-inflammatory and anti-oxidant properties against LPS-induced ALI via inhibiting TLR4 signaling pathway and activating Nrf2 signaling pathway. [ABSTRACT FROM AUTHOR]

Learning from large-scale contrastive language-image pre-training like CLIP has shown remarkable success in a wide range of downstream tasks recently, but it is still under-explored on the challenging few-shot action recognition (FSAR) task. In this work, we aim to transfer the powerful multimodal knowledge of CLIP to alleviate the inaccurate prototype estimation issue due to data scarcity, which is a critical problem in low-shot regimes. To this end, we present a CLIP-guided prototype modulating framework called CLIP-FSAR, which consists of two key components: a video-text contrastive objective and a prototype modulation. Specifically, the former bridges the task discrepancy between CLIP and the few-shot video task by contrasting videos and corresponding class text descriptions. The latter leverages the transferable textual concepts from CLIP to adaptively refine visual prototypes with a temporal Transformer. By this means, CLIP-FSAR can take full advantage of the rich semantic priors in CLIP to obtain reliable prototypes and achieve accurate few-shot classification. Extensive experiments on five commonly used benchmarks demonstrate the effectiveness of our proposed method, and CLIP-FSAR significantly outperforms existing state-of-the-art methods under various settings. The source code and models are publicly available at https://github.com/alibaba-mmai-research/CLIP-FSAR. [ABSTRACT FROM AUTHOR]

In this study, the relationship between the brain parenchymal density, the cerebral vessel density, the mean corpuscular hemoglobin (MCH) content, the mean corpuscular hemoglobin concentration (MCHC), and the morbidity associated with lacunar infarction of residents living in either the plains or the plateau regions were analyzed and compared for their potential clinical implications. Clinical data from the brain CT scans of individuals living in either the plain or plateau regions (129 each) were collected. Specifically, the CT values for basal ganglia, the middle cerebral artery, and the superior sagittal sinus, along with the number of patients with lacunar infarction, were collected. In addition, the MCH and MCHC values were measured in blood samples collected within 48 h following the CT scans. For statistical analysis, an independent sample t-test, Pearson's correlation test (permutation test), and Chi-squared test were employed. The inhabitants of the plateau had a significantly higher CT value of basal ganglia, the middle cerebral artery, and superior sagittal sinus and also higher levels of MCH and MCHC in the blood (ps < 0.001) than the inhabitants of the plains region. Further, there was a significant positive correlation between the three aforementioned CT values and the MCH and MCHC findings. However, no significant differences were found in the morbidity of lacunar infarction between these two regions (p > 0.05). The inhabitants in the plateau have a significantly higher brain parenchymal density, higher CT value for cerebral vessels density, and higher blood MCH and MCHC levels in comparison with individuals occupying the plains. Concurrently, the parenchymal density and the CT values are shown to be positively correlated with the MCH and MCHC content in the blood. [ABSTRACT FROM AUTHOR]

Purpose: The purpose of the study was to examine health-related quality of life (HRQoL) about the most common cancers survivors (lung, stomach, colorectal, breast, and esophageal cancer) in rural China.Methods: We administrated a cross-sectional study in three counties in Shandong province from August to September 2017. The five-level EuroQol-5-dimension (EQ-5D-5L) questionnaire was used to measure the HRQoL among cancer patients at least 8 months post-diagnosis. The Chinese population-based preference trade-off time (TTO) model and discrete choice experiment (DCE) were used to convert the EQ-5D-5L utility score. Tobit regression model was used to identify independent associations between socio-demographic, clinical variables with the HRQoL.Results: In total, 452 cancer survivors were included. The mean EQ-5D-5L utility scores and Visual Analog Scale (EQ-VAS) scores were 0.841 (SD = 0.233) and 70.35 (SD = 18.80) for cancer survivors, respectively. Among the five dimensions, 58.6% of survivors had at least slight levels of pain/discomfort, and 39.2% showed at least slight levels of anxiety/depression. The influencing factors of HRQoL included cancer stage at diagnosis, tumor site, comorbidities, annual household income, and migrant worker status (rural-to-urban migration). Compared to other cancer patients, lung cancer patients had the lowest HRQoL. Higher household income and being a migrant worker were associated with a higher HRQoL for cancer survivors.Conclusions: Cancer survivors in rural China have deteriorated HRQoL, and a substantial number of survivors have pain/discomfort problems. Our study provides detailed data on HRQoL of rural cancer survivors for future supportive and survivorship care in China. [ABSTRACT FROM AUTHOR]

Existing action recognition algorithms require a set of positive exemplars to train a classifier for each action. However, the amount of action classes is very large and the users' queries vary dramatically. It is impractical to pre-define all possible action classes beforehand. To address this issue, we propose to perform action recognition with no positive exemplars, which is often known as the zero-shot learning. Current zero-shot learning paradigms usually train a series of attribute classifiers and then recognize the target actions based on the attribute representation. To ensure the maximum coverage of ad-hoc action classes, the attribute-based approaches require large numbers of reliable and accurate attribute classifiers, which are often unavailable in the real world. In this paper, we propose an approach that merely takes an action name as the input to recognize the action of interest without any pre-trained attribute classifiers and positive exemplars. Given an action name, we first build an analogy pool according to an external ontology, and each action in the analogy pool is related to the target action at different levels. The correlation information inferred from the external ontology may be noisy. We then propose an algorithm, namely adaptive multi-model rank-preserving mapping (AMRM), to train a classifier for action recognition, which is able to evaluate the relatedness of each video in the analogy pool adaptively. As multiple mapping models are employed, our algorithm has better capability to bridge the gap between visual features and the semantic information inferred from the ontology. Extensive experiments demonstrate that our method achieves the promising performance for action recognition only using action names, while no attributes and positive exemplars are available. [ABSTRACT FROM AUTHOR]