Estimating 3D bounding boxes from monocular images is an essential component in autonomous driving, while accurate 3D object detection from this kind of data is very challenging. In this work, by intensive diagnosis experiments, we quantify the impact introduced by each sub-task and found the 'localization error' is the vital factor in restricting monocular 3D detection. Besides, we also investigate the underlying reasons behind localization errors, analyze the issues they might bring, and propose three strategies. First, we revisit the misalignment between the center of the 2D bounding box and the projected center of the 3D object, which is a vital factor leading to low localization accuracy. Second, we observe that accurately localizing distant objects with existing technologies is almost impossible, while those samples will mislead the learned network. To this end, we propose to remove such samples from the training set for improving the overall performance of the detector. Lastly, we also propose a novel 3D IoU oriented loss for the size estimation of the object, which is not affected by 'localization error'. We conduct extensive experiments on the KITTI dataset, where the proposed method achieves real-time detection and outperforms previous methods by a large margin. The code will be made available at: https://github.com/xinzhuma/monodle.
Being the trend of TV developing,DTV is widely concerned in industry and academe.As one of the key technologies of DTV,video encoding becomes the main point.This paper introduces the history of video encoding technology which is going to be used in DTV as well as the principium and characteristics of H.264.
e15541 Background: The role of aldehyde dehydrogenase 2 (ALDH 2) in alcohol metabolism has been ascertained, yet the association between ALDH2 genotype and the carcinogenesis of gastric cancer remains controversy. Methods: We retrospectively collected data of 292 subjects hospitalized in Zhongshan Hospital in Shanghai during March 2016 to September 2016. All subjects were divided into case group (identified with gastric cancer by postoperative pathology) and control group. Distribution of aldehyde dehydrogenase 2 genotype, drinking status along with other clinical variables were compared. Logistic regression model included age, gender, smoking and drinking status, family history and ALDH2 genotype. Adjusted odds ratios (ORs) were calculated. Results: There was a significant elevation of alchohol consumption in case group (P = 0.033). Subjects with ALDH2 *1/*2 and *2/*2 genotype showed less inclination for drinking than ALDH2 *1/*1 (P < 0.001). The risk for gastric cancer showed no differences among subjects with ALDH2 genotype adjusted drinking status [OR 1.121, 95% confidence interval (CI) 0.555-2.263 in Never/rare drinkers; OR 3.380, 95% CI 0.232-49.240 in Light drinkers and OR 2.416, 95% CI 0.326-14.151). Further analysis indicated that high body weight index (BMI) and heavy smoking were individual risk factors for gastric cancer among Light/heavy dringkers with ALDH2 *1/*2 or *2/*2 genotype (OR 84.736, P = 0.021 for BMI and OR 3.904, P = 0.037 for smoking). Conclusions: A dose-response relationship between alchohol consumption and risk for gastric cancer was observed in primary analysis. However ALDH2 genotype failed to modify the alchohol consumption. Our findings suggested that there might be other pathways to further explain the mechanisim of the role of alchohol consumption in carcinogenisis of gastric cancer.
We address the challenges in automatic face recognition (AFR) applications when probe images present multiple variations including pose and resolution changes. Existing approaches attempt to seek a common feature space shared by these variations through linear or local linear mappings. In this paper, we leverage deep learning as a natural feature representation to discover intrinsic nonlinear relationships between images of multiple variations. Our method also extends the locality preserving projection (LPP) with nonlinear mappings learned through optimizing the objective function that preserves local neighboring structures between couterpart images. We perform the experiments on images from several available databases where only one frontal upright image presents in the gallery and variations on pose and resolution appear in the probe. The experiments show the superior recognition rates of our approach over the latest linear (or locally linear) methods.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)