Search

In this paper, a simple yet effective network pruning framework is proposed to simultaneously address the problems of pruning indicator, pruning ratio, and efficiency constraint. This paper argues that the pruning decision should depend on the convolutional weights, and thus proposes novel weight-dependent gates (W-Gates) to learn the information from filter weights and obtain binary gates to prune or keep the filters automatically. To prune the network under efficiency constraints, a switchable Efficiency Module is constructed to predict the hardware latency or FLOPs of candidate pruned networks. Combined with the proposed Efficiency Module, W-Gates can perform filter pruning in an efficiency-aware manner and achieve a compact network with a better accuracy-efficiency trade-off. We have demonstrated the effectiveness of the proposed method on ResNet34, ResNet50, and MobileNet V2, respectively achieving up to 1.33/1.28/1.1 higher Top-1 accuracy with lower hardware latency on ImageNet. Compared with state-of-the-art methods, W-Gates also achieves superior performance. [ABSTRACT FROM AUTHOR]

We present a new learning-based framework S-3D-RCNN that can recover accurate object orientation in SO(3) and simultaneously predict implicit rigid shapes from stereo RGB images. For orientation estimation, in contrast to previous studies that map local appearance to observation angles, we propose a progressive approach by extracting meaningful Intermediate Geometrical Representations (IGRs). This approach features a deep model that transforms perceived intensities from one or two views to object part coordinates to achieve direct egocentric object orientation estimation in the camera coordinate system. To further achieve finer description inside 3D bounding boxes, we investigate the implicit shape estimation problem from stereo images. We model visible object surfaces by designing a point-based representation, augmenting IGRs to explicitly address the unseen surface hallucination problem. Extensive experiments validate the effectiveness of the proposed IGRs, and S-3D-RCNN achieves superior 3D scene understanding performance. We also designed new metrics on the KITTI benchmark for our evaluation of implicit shape estimation. [ABSTRACT FROM AUTHOR]

In this paper, we study 1-bit convolutional neural networks (CNNs), of which both the weights and activations are binary. While being efficient, the lacking of a representational capability and the training difficulty impede 1-bit CNNs from performing as well as real-valued networks. To this end, we propose Bi-Real net with a novel training algorithm to tackle these two challenges. To enhance the representational capability, we propagate the real-valued activations generated by each 1-bit convolution via a parameter-free shortcut. To address the training difficulty, we propose a training algorithm using a tighter approximation to the derivative of the sign function, a magnitude-aware binarization for weight updating, a better initialization method, and a two-step scheme for training a deep network. Experiments on ImageNet show that an 18-layer Bi-Real net with the proposed training algorithm achieves 56.4% top-1 classification accuracy, which is 10% higher than the state-of-the-arts (e.g., XNOR-Net), with a greater memory saving and a lower computational cost. Bi-Real net is also the first to scale up 1-bit CNNs to an ultra-deep network with 152 layers, and achieves 64.5% top-1 accuracy on ImageNet. A 50-layer Bi-Real net shows comparable performance to a real-valued network on the depth estimation task with merely a 0.3% accuracy gap. [ABSTRACT FROM AUTHOR]

Cellulose and pectin are the important components of tobacco (Nicotiana tabacum L.) cell wall, which affect the formation of undesirable compounds. Their contents are closely related to the harmfulness of tobacco. But the simultaneous quantitative analysis of cellulose and pectin is hard to be achieved for traditional analytical methods. A solid-state 13C cross-polarization by multiple contact periods (multiCP) NMR method was developed for the simultaneous quantification of cellulose and pectin in tobacco. The multiCP spectrum at optimal parameters agreed well with the direct polarization (DP) spectrum within one-thirtieth of the measurement time and provided satisfactory signal to noise ratio (SNR). After three simple procedures of sample preparation and spectra deconvolution, simultaneous quantification of cellulose and pectin extracted from tobacco was effectively achieved. Compared with the chemical method, this interesting method was rapid, practicable, and very promising, which provided the technical support for the simultaneous quantification of cell wall substances in biological sample. [Display omitted] • MultiCP NMR for determination of cellulose and pectin in tobacco was introduced. • The optimized spectrum addressed the simultaneous quantification of cellulose and pectin. • The measurement time of quantitative NMR was shortened to 107 min with high SNR. • This interesting solid-state NMR method was rapid, practicable, and very promising. [ABSTRACT FROM AUTHOR]

LEDs with the advantage of high luminous efficacy and long life time show the potential of replacing traditional luminaire. Most commercial white LED light sources use blue or ultraviolet chip coated with emitting phosphor, but the sensitivity and instability of such phosphors has become a big issue. The typical RGB-LED by using individual chips has the problem of spatial separation and insufficient spectral overlap which leads to low CRI. This study suggests a novel and high-efficiency design of fiber optical optocoupler to synthesize four colors emitted by separate LEDs to provide the ideal light sources by adjusting the individual LEDs separately. By choosing different colored light to be synthesized, this optocoupler can be used as light sources which can be highly controlled to offer the best lighting conditions. Compared with other widely used commercial LED sources, this new design of light sources can be used in special experiments which require multi-spectral light. [ABSTRACT FROM AUTHOR]

Background: During the biomass-to-bio-oil conversion process, many studies focus on studying the association between biomass and bio-products using near-infrared spectra (NIR) and chemical analysis methods. However, the characterization of biomass pyrolysis behaviors using thermogravimetric analysis (TGA) with support vector machine (SVM) algorithm has not been reported. In this study, tobacco was chosen as the object for biomass, because the cigarette smoke (including water, tar, and gases) released by tobacco pyrolysis reactions decides the sensory quality, which is similar to biomass as a renewable resource through the pyrolysis process. Results: SVM algorithm has been employed to automatically classify the planting area and growing position of tobacco leaves using thermogravimetric analysis data as the information source for the first time. Eighty-eight single-grade tobacco samples belonging to four grades and eight categories were split into the training, validation, and blind testing sets. Our model showed excellent performances in both the training and validation set as well as in the blind test, with accuracy over 91.67%. Throughout the whole dataset of 88 samples, our model not only provides precise results on the planting area of tobacco leave, but also accurately distinguishes the major grades among the upper, lower, and middle positions. The error only occurs in the classification of subgrades of the middle position. Conclusions: From the case study of tobacco, our results validated the feasibility of using TGA with SVM algorithm as an objective and fast method for auto-classification of tobacco planting area and growing position. In view of the high similarity between tobacco and other biomasses in the compositions and pyrolysis behaviors, this new protocol, which couples the TGA data with SVM algorithm, can potentially be extrapolated to the auto-classification of other biomass types. [ABSTRACT FROM AUTHOR]

Cross-domain visual problems, such as image-to-image translation and domain adaptive object detection, have attracted increasing attentions in the last few years, and also become new rising and challenging directions for the computer vision community. Recently, despite enormous efforts of the field in data collection, there are still few datasets covering the instance-level image-to-image translation and domain adaptive object detection tasks simultaneously. In this work, we introduce a large-scale cross-domain benchmark CDTD (contains 155,529 high-resolution natural images across four different modalities with object bounding box annotations. A summary of the entire dataset is provided in the following sections. Dataset is available at: http://zhiqiangshen.com/projects/INIT/index.html.) for the new instance-level translation and object detection tasks. We provide comprehensive baseline results of the benchmark on both of these two tasks. Moreover, we proposed a novel instance-level image-to-image translation approach called INIT and a gradient detach method for the domain adaptive object detection to harvest and exert dataset's function of the instance level annotations across different domains. [ABSTRACT FROM AUTHOR]

The green strategy of catalytic pyrolysis of biomass to produce valuable fuels or chemicals can effectively alleviate the energy and environmental problems associated with the widespread use of fossil energy. In this study, a catalytic system for the production of furans from pyrolysis of cellulose was reported by using red mud (RM) supported transition metals (Cu/RM, Co/RM, Ni/RM and Mo/RM) as catalysts. The results show that Co/RM has superior catalytic performance for producing furans, while Mo/RM has a weaker effect among the four catalysts. The loading of Mo also increases the peak area of furans and aromatics. Cu/RM could promote the formation of ketones and Ni/RM could further convert furans to aromatics. The effect of in-situ and ex-situ catalytic pyrolysis on the formation of products was investigated. The results show that whether in-situ or ex-situ catalytic pyrolysis, Co/RM has superior catalytic activity for the production of furans. The difference is that the peak area of furans in in-situ pyrolysis is 37.98 × 106 / mg and that in ex-situ pyrolysis is 58.89 × 106 / mg. Moreover, the in-situ coupled with ex-situ pyrolysis further enhance the furans production with 62.35 × 106 / mg. [Display omitted] • A catalytic pyrolysis of cellulose over modified RM catalysts was reported. • The in-situ combined with ex-situ catalytic pyrolysis of cellulose maximizes the yield of furans. • Co/RM has superior performance for producing furans with 62.35 × 106/mg. [ABSTRACT FROM AUTHOR]