Search

In autonomous driving, Vehicle-Infrastructure Cooperative 3D Object Detection (VIC3D) makes use of multi-view cameras from both vehicles and traffic infrastructure, providing a global vantage point with rich semantic context of road conditions beyond a single vehicle viewpoint. Two major challenges prevail in VIC3D: 1) inherent calibration noise when fusing multi-view images, caused by time asynchrony across cameras; 2) information loss when projecting 2D features into 3D space. To address these issues, We propose a novel 3D object detection framework, Vehicles-Infrastructure Multi-view Intermediate fusion (VIMI). First, to fully exploit the holistic perspectives from both vehicles and infrastructure, we propose a Multi-scale Cross Attention (MCA) module that fuses infrastructure and vehicle features on selective multi-scales to correct the calibration noise introduced by camera asynchrony. Then, we design a Camera-aware Channel Masking (CCM) module that uses camera parameters as priors to augment the fused features. We further introduce a Feature Compression (FC) module with channel and spatial compression blocks to reduce the size of transmitted features for enhanced efficiency. Experiments show that VIMI achieves 15.61% overall AP_3D and 21.44% AP_BEV on the new VIC3D dataset, DAIR-V2X-C, significantly outperforming state-of-the-art early fusion and late fusion methods with comparable transmission cost., Comment: 8 pages, 9 figures

In the past years, learned image compression (LIC) has achieved remarkable performance. The recent LIC methods outperform VVC in both PSNR and MS-SSIM. However, the low bit-rate reconstructions of LIC suffer from artifacts such as blurring, color drifting and texture missing. Moreover, those varied artifacts make image quality metrics correlate badly with human perceptual quality. In this paper, we propose PO-ELIC, i.e., Perception-Oriented Efficient Learned Image Coding. To be specific, we adapt ELIC, one of the state-of-the-art LIC models, with adversarial training techniques. We apply a mixture of losses including hinge-form adversarial loss, Charbonnier loss, and style loss, to finetune the model towards better perceptual quality. Experimental results demonstrate that our method achieves comparable perceptual quality with HiFiC with much lower bitrate., Comment: CVPR2022 Workshop, 5-th CLIC Image Compression Track

In recent years, there has been widespread attention drawn to convolutional neural network (CNN) based blind image quality assessment (IQA). A large number of works start by extracting deep features from CNN. Then, those features are processed through spatial average pooling (SAP) and fully connected layers to predict quality. Inspired by full reference IQA and texture features, in this paper, we extend SAP ($1^{st}$ moment) into spatial moment pooling (SMP) by incorporating higher order moments (such as variance, skewness). Moreover, we provide learning friendly normalization to circumvent numerical issue when computing gradients of higher moments. Experimental results suggest that simply upgrading SAP to SMP significantly enhances CNN-based blind IQA methods and achieves state of the art performance., Comment: ICIP 2022

In this paper, we consider the problem of bit allocation in Neural Video Compression (NVC). First, we reveal a fundamental relationship between bit allocation in NVC and Semi-Amortized Variational Inference (SAVI). Specifically, we show that SAVI with GoP (Group-of-Picture)-level likelihood is equivalent to pixel-level bit allocation with precise rate \& quality dependency model. Based on this equivalence, we establish a new paradigm of bit allocation using SAVI. Different from previous bit allocation methods, our approach requires no empirical model and is thus optimal. Moreover, as the original SAVI using gradient ascent only applies to single-level latent, we extend the SAVI to multi-level such as NVC by recursively applying back-propagating through gradient ascent. Finally, we propose a tractable approximation for practical implementation. Our method can be applied to scenarios where performance outweights encoding speed, and serves as an empirical bound on the R-D performance of bit allocation. Experimental results show that current state-of-the-art bit allocation algorithms still have a room of $\approx 0.5$ dB PSNR to improve compared with ours. Code is available at \url{https://github.com/tongdaxu/Bit-Allocation-Using-Optimization}., ICML 2023

In this paper, we investigate the problem of bit allocation in Neural Video Compression (NVC). First, we reveal that a recent bit allocation approach claimed to be optimal is, in fact, sub-optimal due to its implementation. Specifically, we find that its sub-optimality lies in the improper application of semi-amortized variational inference (SAVI) on latent with non-factorized variational posterior. Then, we show that the corrected version of SAVI on non-factorized latent requires recursively applying back-propagating through gradient ascent, based on which we derive the corrected optimal bit allocation algorithm. Due to the computational in-feasibility of the corrected bit allocation, we design an efficient approximation to make it practical. Empirical results show that our proposed correction significantly improves the incorrect bit allocation in terms of R-D performance and bitrate error, and outperforms all other bit allocation methods by a large margin. The source code is provided in the supplementary material.

Recently, deep learning methods have shown promising results in point cloud compression. For octree-based point cloud compression, previous works show that the information of ancestor nodes and sibling nodes are equally important for predicting current node. However, those works either adopt insufficient context or bring intolerable decoding complexity (e.g. >600s). To address this problem, we propose a sufficient yet efficient context model and design an efficient deep learning codec for point clouds. Specifically, we first propose a window-constrained multi-group coding strategy to exploit the autoregressive context while maintaining decoding efficiency. Then, we propose a dual transformer architecture to utilize the dependency of current node on its ancestors and siblings. We also propose a random-masking pre-train method to enhance our model. Experimental results show that our approach achieves state-of-the-art performance for both lossy and lossless point cloud compression. Moreover, our multi-group coding strategy saves 98% decoding time compared with previous octree-based compression method.

High-frequency ultrasound (HFU) is well suited for imaging embryonic mice due to its noninvasive and real-time characteristics. However, manual segmentation of the brain ventricles (BVs) and body requires substantial time and expertise. This work proposes a novel deep learning based end-to-end auto-context refinement framework, consisting of two stages. The first stage produces a low resolution segmentation of the BV and body simultaneously. The resulting probability map for each object (BV or body) is then used to crop a region of interest (ROI) around the target object in both the original image and the probability map to provide context to the refinement segmentation network. Joint training of the two stages provides significant improvement in Dice Similarity Coefficient (DSC) over using only the first stage (0.818 to 0.906 for the BV, and 0.919 to 0.934 for the body). The proposed method significantly reduces the inference time (102.36 to 0.09 s/volume around 1000x faster) while slightly improves the segmentation accuracy over the previous methods using slide-window approaches., Comment: Full Paper Submission to ISBI 2020

Objectives To observe the level of Cystatin C(Cys C) in patients underwent coronary angiography (CAG) and / or percutaneous coronary intervention (PCI) before and after operation and to explore Cys C value in predicting of contrast-induced nephropathy (CIN). To observe the incidence of CIN in three groups: the basic treatment group, the Vitamine C group and the prostaglandin E1 group and to explore the effective prevention method of CIN. Methods All patients underwent CAG and / or PCI were from the department of Cardiology, Affiliated Hospital of Xuzhou Medical College from March 2011 to September 2011. 200 cases were randomly divided into three groups. The basic treatment group (66 cases) were given 0.9% sodium chloride with the speed of 1 ml/(kg·h) 6h before and 12h after operation, atorvastatin 20mg was given before and after operation. The prostaglandin E1 group (67 cases) were given prostaglandin E1 20μg Qd, 1 day before and 2 consecutive days after operation on the basis of the treatment group. The Vitamin C group (67 cases) were given Vitamine C 3.0 g by intravenous infusion 2∼4 h before operation and oral Vitamin C 1.0 g on day 1, 2 respectively after operation on the basis of the treatment group, the total dose of the Vitamine C is 5 g. The renal function of all patients were accessed before and 1-2 days after operation. The eGFR were calculated by the model of MDRD formula modified to suit the operational practice of Chinese people. Cystatin C level was unified to be determined by Enzyme-linked immunosorbent assay (ELISA). Results The groups were well matched for baseline characteristics, the average amount of contrast medium during operation and the operation duration. Of the 200 patients, there were 23 cases of CIN. The incidence of CIN was 16.7% (11/66) in basic treatment group while 3.0% (2/67) in prostaglandin E1 group and 14.9% (10/67) in Vitamine C group. There was a significant difference in the incidence of CIN between three groups ( P There was a significant difference ( P Conclusions Urinary Cystatin C might be a better and earlier biomarker for prediction of CIN. Hydration and statin with prostaglandin E1 is more effective in reducing the prevalence of CIN in patients underwent CAG and/or PCI. Vitamine C besides hydration and statin in the short time can not further reduce the prevalence of CIN.

Objectives The Braunwald classification and TIMI score are the most useful and frequently used tools of clinical assessment for stratifying cardiovascular risk in patients with unstable angina (UA). However, these scores have a limited capacity when they are used alone. This study sought to evaluate the value of multiple blood biomarkers and computed tomographic angiography (CTA) characteristics in patients with UA. Furthermore, to compare the Braunwald classification and TIMI score with the two new created risk scores based on biomarkers and CTA for predicting short-term outcomes in patients with UA. Methods The study group consisted of 186 patients confirmed UA between November 2011 and May 2012. Diagnosis was made by assessing chest pain, typical electrocardiographic changes and/or elevation of cardiac enzymes. Follow-up time was 6 months with major adverse cardiac events (MACEs) as cardiovascular death, revascularization, recurrent acute coronary syndrome (ACS) and re-admission in hospital. Multiple Blood biomarkers including high-sensitivity cardiac troponinT (hs-cTnT), N-terminal pro–B-type natriuretic peptide (NT-proBNP), high-sensitivity C-reactive protein (hs-CRP), myeloperoxidase (MPO) and ischemia-modified albumin (IMA) were measured. CTA characteristics such as stenosis, plaque, epicardial fat volume (EFV) and calcification were evaluated. MACEs and biomarkers were analysed. Also, the relationships between MACEs and CTA characteristics were analysed. The two novel risk scores including SPEC (Stenosis, Plaque, EFV and Calcification by CTA) and BETTER (BiomarkErs and compuTed Tomography scorE on Risk stratification) scores were assessed in 186 patients. ROC (receiver-operating characteristics) curves were used to evaluate the ability of variables and the risk scores including Braunwald, TIMI, SPEC and BETTER to predict MACEs. Results In 6 months follow-up, 107 (57.53%) MACEs occurred, consisting of 2 cardiac deaths (1.07%), 15 non-fatal myocardial infarctions (8.06%), revascularization including 59 PCI (31.72%) and 5 CABG (2.69%), 26 recurrent ACS and re-admission in hospital (13.98%). Serum levels of MPO, NT-proBNP, hs-cTnT, hs-CRP and CTA characteristics of stenosis, plaque, EFV and calcification in the group with MACEs (n = 107) were significantly higher than the group without MACEs (n = 79). However, IMA were no statistics differences between two groups. Area under curves (AUC) of ROC for MPO, NT-proBNP, hs-cTnT, hs-CRP and IMA were 0.614, 0.622, 0.542, 0.626 and 0.579 respectively. Area under curves (AUC) of ROC for stenosis, plaque, EFV and calcification were 0.834, 0.766, 0.666 and 0.633 respectively. The AUC for SPEC was greater than Braunwald classification and TIMI risk scores (0.902 vs 0.591 and 0.643, p 0.05). Conclusions The SPEC risk score which based on CTA and BETTER risk scores integrating CTA and biomarkers are both new tools specifically developed for patients with UA. These scores displayed prediction accuracy in terms of discrimination and calibration than other currently available scores for risk stratification.

Objective The aim of the present study was to assess the influence of anaemia on development of contrast-induced acute kidney injury (AKI) after percutaneous coronary intervention. Methods The subject group consisted of 1026 patients who had undergone coronary intervention procedure between January 1, 2008 and October 31, 2009. A nonionic, low osmolality contrast agent was used in our laboratory at this time. Serum creatinine values were measured before and within 48 h of administration of contrast agents. Contrast-induced nephropathy was defined as an increase of 3 0.5 mg/dl or 3 25% in serum creatinine concentration over baseline within 48 h of angiography, Anaemia as haemoglobin (Hgb) 120 g/l in women and 130 g/l in men. Results Among the 1026 patients studied, 32 (3.1%) experienced AKI after procedure. AKI occurred in 6.3% of the anaemic patients and 2.2% of the non-anaemic patients (p 0.05). Multivariate logistic regression analysis found that baseline creatinine clearance and baseline haemoglobin were independent predictors of AKI. When presence of anaemia was introduced into the multivariate model instead of baseline haemoglobin, it was also showed a significant association with AKI. Conclusion Anemia increases the incidence of AKI in patients with moderate renal dysfunction. Patients with both preexisting renal insufficiency and anaemia are at high risk of AKI. Baseline creatinine clearance and baseline haemoglobin (or anaemia) are independent predictors of AKI.

Objective To observe the level of neutrophil gelatinase-associated lipocalin (NGAL) in patients undergoing coronary angiography (CAG) and /or percutaneous coronary intervention (PCI) procedure before and after operation and to explore NGAL value in predicting of contrast- induced acute renal injury (CI-AKI). Methods 60 patients undergoing CAG and /or PCI were enrolled in this study, Serum creatinine (Scr) and urinary NGAL were detected at different time points before and after operation. Results Of all 60 patients, the change of the Scr level were not statistically significant (p>0.05), there were no case of CI-AKI. The level of urine NGAL at 6, 12, 24 h after operation was significantly increased compared with baseline (p Conclusion Urinary NGAL has been significantly increased after coronary interventional therapy without the change of Scr levels. Urinary NGAL might be an early biomarker of prediction CI-AKI better than Scr.