Automatic Transformation Search Against Deep Leakage From Gradients.

Collaborative learning has gained great popularity due to its benefit of data privacy protection: participants can jointly train a Deep Learning model without sharing their training sets. However, recent works discovered that an adversary can fully recover the sensitive training samples from the shared gradients. Such reconstruction attacks pose severe threats to collaborative learning. Hence, effective mitigation solutions are urgently desired. In this paper, we systematically analyze existing reconstruction attacks and propose to leverage data augmentation to defeat these attacks: by preprocessing sensitive images with carefully-selected transformation policies, it becomes infeasible for the adversary to extract training samples from the corresponding gradients. We first design two new metrics to quantify the impacts of transformations on data privacy and model usability. With the two metrics, we design a novel search method to automatically discover qualified policies from a given data augmentation library. Our defense method can be further combined with existing collaborative training systems without modifying the training protocols. We conduct comprehensive experiments on various system settings. Evaluation results demonstrate that the policies discovered by our method can defeat state-of-the-art reconstruction attacks in collaborative learning, with high efficiency and negligible impact on the model performance.

Extrinsic calibration for multi-LiDAR systems involving heterogeneous laser scanning models.

In the realm of autonomous driving, there is a pressing demand for heightened perceptual capabilities, giving rise to a plethora of multisensory solutions. Among these, multi-LiDAR systems have gained significant popularity. Within the spectrum of available combinations, the integration of repetitive and non-repetitive LiDAR configurations emerges as a balanced approach, offering a favorable trade-off between sensing range and cost. However, the calibration of such systems remains a challenge due to the diverse nature of point clouds, low-common-view, and distinct densities. This study proposed a novel targetless calibration algorithm for extrinsic calibration between Hybrid-Solid-State-LiDAR(SSL) and Mechanical-LiDAR systems, each employing different scanning modes. The algorithm harnesses planar features within the scene to construct matching costs, while proposing the adoption of the Gaussian Mixture Model (GMM) to address outliers, thereby mitigating the issue of overlapping points. Dynamic trust-region-based optimization is incorporated during iterative processes to enhance nonlinear convergence speed. Comprehensive evaluations across diverse simulated and real-world scenarios affirm the robustness and precision of our algorithm, outperforming current state-of-the-art methods.

Laser reflectance feature assisted accurate extrinsic calibration for non-repetitive scanning LiDAR and camera systems.

Non-repetitive scanning Light Detection And Ranging(LiDAR)-Camera systems are commonly used in autonomous navigation industries, benefiting from their low-cost and high-perception characteristics. However, due to the irregular scanning pattern of LiDAR, feature extraction on point cloud encounters the problem of non-uniformity distribution of density and reflectance intensity, accurate extrinsic calibration remains a challenging task. To solve this problem, this paper presented an open-source calibration method using only a printed chessboard. We designed a two-stage coarse-to-fine pipeline for 3D corner extraction. Firstly, a Gaussian Mixture Model(GMM)-based intensity cluster approach is proposed to adaptively identify point segments in different color blocks of the chessboard. Secondly, a novel Iterative Lowest-cost Pose(ILP) algorithm is designed to fit the chessboard grid and refine the 3D corner iteratively. This scheme is unique for turning the corner feature extraction problem into a grid align problem. After the corresponding 3D-2D points are solved, by applying the PnP(Perspective-n-Point) method, along with nonlinear-optimization refinement, the extrinsic parameters are obtained. Extensive simulation and real-world experimental results show that our method achieved subpixel-level precision in terms of reprojection error. The comparison demonstrated that the effectiveness and accuracy of the proposed method outperformed existing methods.

Qualitative and Quantitative Analysis of Ejiao-Related Animal Gelatins through Peptide Markers Using LC-QTOF-MS/MS and Scheduled Multiple Reaction Monitoring (MRM) by LC-QQQ-MS/MS.

Donkey-hide gelatin, also called Ejiao (colla corii asini), is commonly used as a food health supplement and valuable Chinese medicine. Its growing popular demand and short supply make it a target for fraud, and many other animal gelatins can be found as adulterants. Authentication remains a quality concern. Peptide markers were developed by searching the protein database. However, donkeys and horses share the same database, and there is no specific marker for donkeys. Here, solutions are sought following a database-independent strategy. The peptide profiles of authentic samples of different animal gelatins were compared using LC-QTOF-MS/MS. Fourteen specific markers, including four donkey-specific, one horse-specific, three cattle-specific, and six pig-specific peptides, were successfully found. As these donkey-specific peptides are not included in the current proteomics database, their sequences were determined by de novo sequencing. A quantitative LC-QQQ multiple reaction monitoring (MRM) method was further developed to achieve highly sensitive and selective analysis. The specificity and applicability of these markers were confirmed by testing multiple authentic samples and 110 batches of commercial Ejiao products, 57 of which were found to be unqualified. These results suggest that these markers are specific and accurate for authentication purposes.

Feature homogenization strategy using optical flow for vision-related simultaneous localization and mapping in a complex environment.

Vision-related state estimation usually extracts multiple feature points from images captured by the camera. In this paper, we propose a robust feature homogenization method for resolving the problem of feature clustering. The proposed method deduced the depth of feature points from optical flow magnitude, and the homogenization of feature points was acquired by adaptively enforcing the minimum distance between neighboring feature points. With the assistance of optical flow, the proposed method develops a preference for feature points with smaller depths in feature homogenization. Experimental results show that the proposed method improves the system's global consistency and tracking stability by using optical flow information.

Target classification of multislit streak tube imaging lidar based on deep learning.

To reduce the impact of the image reconstruction process and improve the identification efficiency of the multislit streak tube imaging lidar (MS-STIL) system, an object classification method based on the echo of the MS-STIL system is proposed. A streak image data set is constructed that contains a total of 240 common outdoor targets in 6 categories. Additionally, the deep-learning network model based on ResNet is chosen to implement streak image classification. The effects of two classification methods based on streak images and reconstructed depth images are compared. To verify the maximum classification capability of the proposed method, the recognition effects are investigated under 6 and 20 classes. The results show that the classification accuracy decreases from 99.42% to 67.64%. After the data set is expanded, the classification accuracy improved to 85.35% when the class number of the target is 20.

Denoising of the multi-slit streak tube imaging LiDAR system using a faster non-local mean method.

Combined with the characteristic of the multi-slit streak tube imaging LiDAR (MS-STIL) system where the imaging areas corresponding to each slit do not interfere with each other, we denoised the streak images by an improved fast non-local mean filtering algorithm. Experiments were performed to investigate the effectiveness of the method. The experimental results show that the spatial resolution of the system is improved from 22 to 16 mm; the relative distance error is reduced by an average of 22.76%; and the intensity accuracy improved significantly when the distance is 10 m. Additionally, the overall denoising effect is comprehensively verified by long-range target imaging. The mean square error of the reconstructed depth image and intensity image are reduced from 0.0836, 0.0067 to 0.0433, 0.0037, respectively. The applicability of the proposed method was verified through comparative experiments in different environments.

N-Doped Mesoporous Carbons: From Synthesis to Applications as Metal-Free Reduction Catalysts and Energy Storage Materials.

N-doped mesoporous carbons, NMCs, have attracted intensive attention recently and have shown potential applications in various scientific fields including catalysis and energy conversion/storage. Via modification with foreign N elements and construction of mesoporous structures for NMCs, their electronic and spin structure, as well as their porosity can be greatly tailored. And the resultant electron-donor property, surface wettability, conductivity, ion/molecular transfer and reactivity are changed accordingly. In this review, we will summarize the recent research progress of these metal-free NMCs, with an emphasis on their synthesis and performance, especially for their synthetic strategy and catalytic properties toward oxygen and nitro compound reductions, as well as their electrochemical properties as electrode materials for lithium-ion/sulfur batteries and supercapacitors. We hope for future developments, such as controlling doping methods more precisely, generating more active sites by N-doping, and finding wider applications of NMCs in other fields.

Improving antioxidant and antiproliferative activities of colla corii asini hydrolysates using ginkgo biloba extracts.

Colla corii asini hydrolysates (ACCH) and ginkgo biloba extracts (EGb) possess more potent antioxidant effects when used in combination than when used alone. The mixture of ACCH and EGb at a dose ratio of 20:4(w:w) showed the highest radical scavenging activity with IC 50 of 0.17 ± 0.01, 0.43 ± 0.02 and 1.52 ± 0.07 mg/ml against DPPH, ABTS and HO · free radicals, respectively. Furthermore, the inhibition of breast cancer cells MCF-7 and MDA-MB-231 proliferation increased when these cell lines were treated with a combination of ACCH and EGb for 72 hr, with IC 50 of 4.32 ± 0.12 mg/ml and 0.39 ± 0.01 mg/ml, respectively. The findings indicated that the mixtures of ACCH and EGb could be used to prevent and treat some diseases caused by the excessive free radicals, especially cancer. Therefore, the mixtures of ACCH and EGb might serve as a natural source of desirable antioxidant and anticancer agents for the nutraceutical and pharmaceutical industries.

A rapid and simple UPLC-MS/MS method using collagen marker peptides for identification of porcine gelatin.

Gelatin, which is mainly derived from bovine and porcine sources, has been used in many foods and pharmaceutical products. To ensure the compliance of food products with halal regulations, reliable analytical methods are very much required. In this study, one unique marker peptide for porcine gelatins was selected to develop an UPLC-MS/MS multiple reaction monitoring method. The capability of this method to identify porcine materials was demonstrated by analyzing in-house-made gelatins containing different amounts of porcine gelatins and commercial gelatin products. The adulteration of porcine gelatin could be sensitively detected at a low level of 0.04%. When combined with HPLC and mass spectrometry, this method is an accurate and sensitive quantitative method to identify porcine gelatins. Thus, the strategy can be used to verify halal authenticity of gelatin.

Fast-track surgery decreases the incidence of postoperative delirium and other complications in elderly patients with colorectal carcinoma.

OBJECTIVE: This study aims to investigate the role of fast-track surgery in preventing the development of postoperative delirium and other complications in elderly patients with colorectal carcinoma. METHODS: A total of 240 elderly patients with colorectal carcinoma (aged ≥70 years) undergoing open colorectal surgery was randomly assigned into two groups, in which the patients were managed perioperatively either with traditional or fast-track approaches. The length of hospital stay (LOS) and time to pass flatus were compared. The incidence of postoperative delirium and other complications were evaluated. Serum interleukin-6 (IL-6) levels were determined before and after surgery. RESULTS: The LOS was significantly shorter in the fast-track surgery (FTS) group than that in the traditional group. The recovery of bowel movement (as indicated by the time to pass flatus) was faster in the FTS group. The postoperative complications including pulmonary infection, urinary infection and heart failure were significantly less frequent in the FTS group. Notably, the incidence of postoperative delirium was significantly lower in patients with the fast track therapy (4/117, 3.4 %) than with the traditional therapy (15/116, 12.9 %; p = 0.008). The serum IL-6 levels on postoperative days 1, 2, and 3 in patients with the fast-track therapy were significantly lower than those with the traditional therapy (p < 0.001). CONCLUSIONS: Compared to traditional perioperative management, fast-track surgery decreases the LOS, facilitates the recovery of bowel movement, and reduces occurrence of postoperative delirium and other complications in elderly patients with colorectal carcinoma. The lower incidence of delirium is at least partly attributable to the reduced systemic inflammatory response mediated by IL-6.