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.

To meet the demand of modern acoustic absorbing material for which acoustic absorbing frequency region can be readily tailored, we introduced woodpile structure into locally resonant phononic crystal ͑LRPC͒ and fabricated an underwater acoustic absorbing material, which is called locally resonant phononic woodpile ͑LRPW͒. Experimental results show that LRPW has a strong capability of absorbing sound in a wide frequency range. Further theoretical research revealed that LRPC units and woodpile structure in LRPW play an important role in realization of wide band underwater strong acoustic absorption.

No abstract

No abstract

Micro-Raman spectroscopic analysis of microscopically distinguishable components in a series of high-rank coals (anthracite to graphitized coal) adjacent to a granitic pluton was used to assess the structural evolution of coal during natural graphitization. Microscopically identifiable components were differentiated into six groups: vitrinite; inertinite; microcrystalline graphite with a fine, granular texture and a low reflectance; pyrolytic carbon with layering normal to particle edges; and needle graphite and flake graphite, both of which are similar to commercial synthetic graphite. Approaching the intrusion, Raman spectra exhibit a distinctly different evolution for vitrinite and microcrystalline graphite: the D1 band of the first-order Raman spectrum becomes narrower and more intense for vitrinite, whereas the D1 band intensity decreases for the granular, microcrystalline graphite. A plot of full width at half-maximum for the D1 band versus R 1 (intensity ratio of the D1 to the G band) indicates that structural evolution of vitrinite occurs during carbonization, whereas that of the microcrystalline graphite components occurs during graphitization. The increase in the intensity of the 2D1 band and the appearance of the 2450 cm −1 band in the second-order Raman spectrum for the microcrystalline graphite components also suggest that they have reached the graphitization stage. Structural heterogeneity in the metamorphosed coals initially decreases with increased coal rank, but then increases when fine granular particles (microcrystalline graphite) are seen in the highly graphitized coals. The structural heterogeneity of the most graphitized coals increases due to the formation of new components (pyrolytic carbon, and needle and flake graphite). Insights on the structural features and evolution of natural graphitized coals at a maceral scale presented here may be important in future applications, including the production of synthetic graphite from coal and coaly microcrystalline graphite.