Recently, significant improvement has been made on semantic object segmentation due to the development of deep convolutional neural networks (DCNNs). Training such a DCNN usually relies on a large ...number of images with pixel-level segmentation masks, and annotating these images is very costly in terms of both finance and human effort. In this paper, we propose a simple to complex (STC) framework in which only image-level annotations are utilized to learn DCNNs for semantic segmentation. Specifically, we first train an initial segmentation network called Initial-DCNN with the saliency maps of simple images (i.e., those with a single category of major object(s) and clean background). These saliency maps can be automatically obtained by existing bottom-up salient object detection techniques, where no supervision information is needed. Then, a better network called Enhanced-DCNN is learned with supervision from the predicted segmentation masks of simple images based on the Initial-DCNN as well as the image-level annotations. Finally, more pixel-level segmentation masks of complex images (two or more categories of objects with cluttered background), which are inferred by using Enhanced-DCNN and image-level annotations, are utilized as the supervision information to learn the Powerful-DCNN for semantic segmentation. Our method utilizes 40K simple images from Flickr.com and 10K complex images from PASCAL VOC for step-wisely boosting the segmentation network. Extensive experimental results on PASCAL VOC 2012 segmentation benchmark well demonstrate the superiority of the proposed STC framework compared with other state-of-the-arts.
3D spatial information is known to be beneficial to the semantic segmentation task. Most existing methods take 3D spatial data as an additional input, leading to a two-stream segmentation network ...that processes RGB and 3D spatial information separately. This solution greatly increases the inference time and severely limits its scope for real-time applications. To solve this problem, we propose Spatial information guided Convolution (S-Conv), which allows efficient RGB feature and 3D spatial information integration. S-Conv is competent to infer the sampling offset of the convolution kernel guided by the 3D spatial information, helping the convolutional layer adjust the receptive field and adapt to geometric transformations. S-Conv also incorporates geometric information into the feature learning process by generating spatially adaptive convolutional weights. The capability of perceiving geometry is largely enhanced without much affecting the amount of parameters and computational cost. Based on S-Conv, we further design a semantic segmentation network, called Spatial information Guided convolutional Network (SGNet), resulting in real-time inference and state-of-the-art performance on NYUDv2 and SUNRGBD datasets.
Abstract Systemic chemotherapy given at maximum tolerated doses (MTD) has been the mainstay of cancer treatment for more than half a century. In some chemosensitive diseases such as hematologic ...malignancies and solid tumors, MTD has led to complete remission and even cure. The combination of maintenance therapy and standard MTD also can generate good disease control; however, resistance to chemotherapy and disease metastasis still remain major obstacles to successful cancer treatment in the majority of advanced tumors. Metronomic chemotherapy, defined as frequent administration of chemotherapeutic agents at a non-toxic dose without extended rest periods, was originally designed to overcome drug resistance by shifting the therapeutic target from tumor cells to tumor endothelial cells. Metronomic chemotherapy also exerts anti-tumor effects on the immune system (immunomodulation) and tumor cells. The goal of immunotherapy is to enhance host anti-tumor immunities. Adding immunomodulators such as metronomic chemotherapy to immunotherapy can improve the clinical outcomes in a synergistic manner. Here, we review the anti-tumor mechanisms of metronomic chemotherapy and the preliminary research addressing the combination of immunotherapy and metronomic chemotherapy for cancer treatment in animal models and in clinical setting.
Next‐generation batteries based on conversion reactions, including aqueous metal–air batteries, nonaqueous alkali metal‐O2 and ‐CO2 batteries, alkali metal‐chalcogen batteries, and alkali metal‐ion ...batteries have attracted great interest. However, their use is restricted by inefficient reversible conversion of active agents. Developing bifunctional catalysts to accelerate the conversion reaction kinetics in both discharge and charge processes is urgently needed. Graphene‐, or graphene‐like carbon‐supported atomically dispersed metal catalysts (G‐ADMCs) have been demonstrated to show excellent activity in various electrocatalytic reactions, making them promising candidates. Different from G‐ADMCs for catalysis, which only require high activity in one direction, G‐ADMCs for rechargeable batteries should provide high activity in both discharging and charging. This review provides guidance for the design and fabrication of bifunctional G‐ADMCs for next‐generation rechargeable batteries based on conversion reactions. The key challenges that prevent their reversible conversion, the origin of the activity of bifunctional G‐ADMCs, and the current design principles of bifunctional G‐ADMCs for highly reversible conversion, have been analyzed and highlighted for each conversion‐type battery. Finally, a summary and outlook on the development of bifunctional G‐ADMC materials for next‐generation batteries with a high energy density and excellent energy efficiency are given.
This review analyzes the key factors that hinder the reversible conversion of conversion‐type materials, provides a fundamental understanding of graphene‐, or graphene‐like carbon‐supported atomically dispersed metal catalysts (G‐ADMCs), and provides guidance on the bifunctional G‐ADMCs to be used in high‐performance next‐generation batteries, including aqueous metal–air batteries, nonaqueous alkali metal‐O2 and ‐CO2 batteries, alkali metal‐chalcogen batteries, and alkali metal‐ion batteries.
Poly(ethylene oxide) (PEO)‐based electrolytes are promising for all‐solid‐state batteries but can only be used above room temperature due to the high‐degree crystallization of PEO and the intimate ...affinity between ethylene oxide (EO) chains and lithium ions. Here, a homogeneous‐inspired design of PEO‐based solid‐state electrolytes with fast ion conduction is proposed. The homogeneous PEO‐based solid‐state electrolyte with an adjusted succinonitrile (SN) and PEO molar ratio simultaneously suppresses the PEO crystallization and mitigates the affinity between EO and Li+. By adjusting the molar ratio of SN to PEO (SN:EO ≈ 1:4), channels providing fast Li+ transport are formed within the homogeneous solid‐state polymer electrolyte, which increases the ionic conductivity by 100 times and enables their application at a low temperature (0–25 °C), together with the uniform lithium deposition. This modified PEO‐based electrolyte also enables a LiFePO4 cathode to achieve a superior Coulombic efficiency (>99%) and have a long life (>750 cycles) at room temperature. Moreover, even at a low temperature of 0 °C, 82% of its room‐temperature capacity remains, demonstrating the great potential of this electrolyte for practical solid‐state lithium battery applications.
Homogeneous‐inspired design of solid‐state polymer electrolytes with fast ion conduction is proposed. By adjusting the molar ratio of succinonitrile to poly(ethylene oxide) (SN:EO≈1:4), channels providing fast Li+ transport are formed within the homogeneous solid‐state polymer electrolyte, which increases the ionic conductivity by 100 times and enables their application at a low temperature (0–25 °C).
Scaling up to a large number of qubits with high-precision control is essential in the demonstrations of quantum computational advantage to exponentially outpace the classical hardware and ...algorithmic improvements. Here, we develop a two-dimensional programmable superconducting quantum processor, Zuchongzhi, which is composed of 66 functional qubits in a tunable coupling architecture. To characterize the performance of the whole system, we perform random quantum circuits sampling for benchmarking, up to a system size of 56 qubits and 20 cycles. The computational cost of the classical simulation of this task is estimated to be 2–3 orders of magnitude higher than the previous work on 53-qubit Sycamore processor Nature 574, 505 (2019). We estimate that the sampling task finished by Zuchongzhi in about 1.2 h will take the most powerful supercomputer at least 8 yr. Our work establishes an unambiguous quantum computational advantage that is infeasible for classical computation in a reasonable amount of time. The high-precision and programmable quantum computing platform opens a new door to explore novel many-body phenomena and implement complex quantum algorithms.
Advanced electrochemical energy storage devices (EESDs) that can store electrical energy efficiently while being miniature/flexible/wearable/load-bearing are much needed for various applications ...ranging from flexible/wearable/portable electronics to lightweight electric vehicles/aerospace equipment. Carbon-based fibers hold great promise in the development of these advanced EESDs (e.g., supercapacitors and batteries) due to their being lightweight, high electrical conductivity, excellent mechanical strength, flexibility, and tunable electrochemical performance. This review summarizes the fabrication techniques of carbon-based fibers, especially carbon nanofibers, carbon-nanotube-based fibers, and graphene-based fibers, and various strategies for improving their mechanical, electrical, and electrochemical performance. The design, assembly, and potential applications of advanced EESDs from these carbon-based fibers are highlighted. Finally, the challenges and future opportunities of carbon-based fibers for advanced EESDs are discussed.
Increasing fresh water demand for drinking and agriculture is one of the grand challenges of our age. Graphene oxide (GO) membranes have shown a great potential for desalination and water ...purification. However, it is challenging to further improve the water permeability without sacrificing the separation efficiency, and the GO membranes are easily delaminated in aqueous solutions within few hours. Here, we report a class of reduced GO membranes with enlarged interlayer distance fabricated by using theanine amino acid and tannic acid as reducing agent and cross-linker. Such membranes show water permeance over 10,000 L m
h
bar
, which is 10-1000 times higher than those of previously reported GO-based membranes and commercial membranes, and good separation efficiency, e.g., rhodamine B and methylene blue rejection of ~100%. Moreover, they show no damage or delamination in water, acid, and basic solutions even after months.
The challenges of developing neuromorphic vision systems inspired by the human eye come not only from how to recreate the flexibility, sophistication, and adaptability of animal systems, but also how ...to do so with computational efficiency and elegance. Similar to biological systems, these neuromorphic circuits integrate functions of image sensing, memory and processing into the device, and process continuous analog brightness signal in real-time. High-integration, flexibility and ultra-sensitivity are essential for practical artificial vision systems that attempt to emulate biological processing. Here, we present a flexible optoelectronic sensor array of 1024 pixels using a combination of carbon nanotubes and perovskite quantum dots as active materials for an efficient neuromorphic vision system. The device has an extraordinary sensitivity to light with a responsivity of 5.1 × 10
A/W and a specific detectivity of 2 × 10
Jones, and demonstrates neuromorphic reinforcement learning by training the sensor array with a weak light pulse of 1 μW/cm
.
Lithium–sulfur (Li–S) batteries are highly appealing for next‐generation electrochemical energy storage owing to their high theoretical energy density, environmental friendliness, and low cost. ...However, the insulating nature of sulfur and migration of dissolved polysulfide intermediates lead to low active material utilization and fast capacity decay, which pose a significant challenge to their practical applications. Here, this paper reports a multifunctional carbon hybrid with metal–organic frameworks (MOFs)‐derived nitrogen‐doped porous carbon anchored on graphene sheets (NPC/G) serving as a sulfur host. On the one hand, the high surface area and nitrogen‐doping of the carbon nanoparticles enable effective polysulfide immobilization through both physical confinement and chemical adsorption; on the other hand, the highly conductive graphene provides an interconnected conductive framework to facilitate fast electron transport, improving the sulfur utilization. As a result, the NPC/G‐based sulfur cathode exhibits a high specific capacity of 1372 mAh g−1 with good cycling stability over 300 cycles. This approach provides a promising approach for the design of MOFs‐derived carbon materials for high performance Li–S batteries.
A multifunctional carbon hybrid with metal–organic frameworks‐derived nitrogen‐doped porous carbon in situ formed on graphene sheets is prepared for sulfur accommodation. Benefiting from the high conductivity, abundant pore structure and nitrogen doping of the carbon hybrid, the as‐obtained sulfur electrode shows excellent electrochemical performance with a high specific capacity of 1372 mAh g−1 and good cycling stability over 300 cycles.