Search

Image captioning (IC) systems aim to generate a text description of the salient objects in an image. In recent years, IC systems have been increasingly integrated into our daily lives, such as assistance for visually-impaired people and description generation in Microsoft Powerpoint. However, even the cutting-edge IC systems (e.g., Microsoft Azure Cognitive Services) and algorithms (e.g., OFA) could produce erroneous captions, leading to incorrect captioning of important objects, misunderstanding, and threats to personal safety. The existing testing approaches either fail to handle the complex form of IC system output (i.e., sentences in natural language) or generate unnatural images as test cases. To address these problems, we introduce Recursive Object MElting (Rome), a novel metamorphic testing approach for validating IC systems. Different from existing approaches that generate test cases by inserting objects, which easily make the generated images unnatural, Rome melts (i.e., remove and inpaint) objects. Rome assumes that the object set in the caption of an image includes the object set in the caption of a generated image after object melting. Given an image, Rome can recursively remove its objects to generate different pairs of images. We use Rome to test one widely-adopted image captioning API and four state-of-the-art (SOTA) algorithms. The results show that the test cases generated by Rome look much more natural than the SOTA IC testing approach and they achieve comparable naturalness to the original images. Meanwhile, by generating test pairs using 226 seed images, Rome reports a total of 9,121 erroneous issues with high precision (86.47%-92.17%). In addition, we further utilize the test cases generated by Rome to retrain the Oscar, which improves its performance across multiple evaluation metrics., Accepted by ISSTA 2023

Image captioning (IC) systems, which automatically generate a text description of the salient objects in an image (real or synthetic), have seen great progress over the past few years due to the development of deep neural networks. IC plays an indispensable role in human society, for example, labeling massive photos for scientific studies and assisting visually-impaired people in perceiving the world. However, even the top-notch IC systems, such as Microsoft Azure Cognitive Services and IBM Image Caption Generator, may return incorrect results, leading to the omission of important objects, deep misunderstanding, and threats to personal safety. To address this problem, we propose MetaIC, the \textit{first} metamorphic testing approach to validate IC systems. Our core idea is that the object names should exhibit directional changes after object insertion. Specifically, MetaIC (1) extracts objects from existing images to construct an object corpus; (2) inserts an object into an image via novel object resizing and location tuning algorithms; and (3) reports image pairs whose captions do not exhibit differences in an expected way. In our evaluation, we use MetaIC to test one widely-adopted image captioning API and five state-of-the-art (SOTA) image captioning models. Using 1,000 seeds, MetaIC successfully reports 16,825 erroneous issues with high precision (84.9\%-98.4\%). There are three kinds of errors: misclassification, omission, and incorrect quantity. We visualize the errors reported by MetaIC, which shows that flexible overlapping setting facilitates IC testing by increasing and diversifying the reported errors. In addition, MetaIC can be further generalized to detect label errors in the training dataset, which has successfully detected 151 incorrect labels in MS COCO Caption, a standard dataset in image captioning.

We demonstrated multi-mobile robot navigation based on Visible Light Positioning(VLP) localization. From our experiment, the VLP can accurately locate robots' positions in navigation.

Zhiqing Zhong,1,* Hongyan Li,2,* Hongzhen Zhong,1 Tianbiao Zhou1 1Department of Nephrology, The Second Affiliated Hospital, Shantou University Medical College, 515041 Shantou, China; 2Department of Nephrology, Huadu District People’s Hospital of Guangzhou, Southern Medical University, 510800 Guangzhou, China *These authors contributed equally to this work Background: Long-term treatment programs with low toxicity represent a therapeutic challenge in lupus nephritis (LN). Although a therapeutic benefit of rituximab (RTX) has been reported in LN patients who have failed conventional treatment, the results are controversial. We aimed to assess the clinical efficacy and safety of RTX as a new immunosuppressive medicine in the treatment of LN with a meta-analysis. Methods: Based on predetermined criteria, PubMed, Embase, and Cochrane Library were used to identify the eligible studies. Cochrane Review Manager version 5.3 was applied to pool the data extracted from individual investigations and provide summary effect estimates. Results: Twenty-four studies with 940 patients were analyzed. In case series trials with specific LN assessment, the complete remission (CR) rate at 12 months was 35.9% (95% CI: 24.2%–49.5%), and total remission (TR: CR plus partial remission) was 73.4% (95% CI: 66.0%–79.7%). In controlled trials, RTX was associated with a higher probability of TR (OR =2.02, 95% CI: 1.23–3.32, P0.05). Additionally, RTX treatment significantly decreased proteinuria (mean difference: -2.79, 95% CI: -3.95 to -1.62, P

As the non-spherical particles could be distinguished from spherical particles easily from polarization lidar measurement, polarization lidar was widely employed in the atmospheric measurements. However, it is very difficult to retrial the particle properties quantitatively from depolarization measurement due to the complex vector scattering model and the calibration of depolarization measurement. In this paper, a method for depolarization calibration of polarization lidar was introduced, and the offset angle of the half-wave plate and gain ratio could be determined exactly by using fitting method. The simulation and lidar measurements indicate that the polarization lidar could be calibrated well by using this method.

The Raman lidar has been developed and installed in the Beijing from November to December in 2014 for the observation of the spatial and vertical characteristics of aerosol and water vapor in the planetary boundary layer. During the experiments, the Raman lidar operates in the automatic and continuous mode with the spatial resolution of 7.5 m and the temporal resolution of 15 min. The observation results show that the aerosol extinction always increases along with the greater content of water vapor and the lower height of planetary boundary layer. It implies that the decrease of air quality in Beijing is associated with the water vapor in the air and the height of planetary boundary layer beside the enrichment of local and transported aerosols.

A high spectral resolution lidar(HSRL) based on injection-seeded laser and iodine absorption filter has been developed. The basic principle, data processing method and configuration of HSRL were introduced. The HSRL measurement of atmospheric transmission in cloud and dust cases were presented,and compared with the results retrieved by the traditional Fernald method. Due to the nonidentity in physical and optical properties of different aerosols and cloud, the lidar ratio assumed in Fernald method cannot agrees well with the real atmosphere and produces considerable error in atmospheric transmission. Utilizing different spectral width of molecular scattering and particle scattering, the HSRL could separate two kinds of signals with iodine filter. Therefore, HSRL can measure the atmospheric transmission quantitatively without the additional assumption of lidar ratio.