Shenzhen Institutes of Advanced Technology

Motivation: Quality control and preprocessing of FASTQ files are essential to providing clean data for downstream analysis. Traditionally, a different tool is used for each operation, such as quality control, adapter trimming and quality filtering. These tools are often insufficiently fast as most are developed using high-level programming languages (e.g. Python and Java) and provide limited multi-threading support. Reading and loading data multiple times also renders preprocessing slow and I/O inefficient. Results: We developed fastp as an ultra-fast FASTQ preprocessor with useful quality control and data-filtering features. It can perform quality control, adapter trimming, quality filtering, per-read quality pruning and many other operations with a single scan of the FASTQ data. This tool is developed in C++ and has multi-threading support. Based on our evaluation, fastp is 2-5 times faster than other FASTQ preprocessing tools such as Trimmomatic or Cutadapt despite performing far more operations than similar tools. Availability and implementation: The open-source code and corresponding instructions are available at https://github.com/OpenGene/fastp.

Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts.

AUTORES: Daniel J Klionsky1745,1749*, Kotb Abdelmohsen840, Akihisa Abe1237, Md Joynal Abedin1762, Hagai Abeliovich425,
\nAbraham Acevedo Arozena789, Hiroaki Adachi1800, Christopher M Adams1669, Peter D Adams57, Khosrow Adeli1981,
\nPeter J Adhihetty1625, Sharon G Adler700, Galila Agam67, Rajesh Agarwal1587, Manish K Aghi1537, Maria Agnello1826,
\nPatrizia Agostinis664, Patricia V Aguilar1960, Julio Aguirre-Ghiso784,786, Edoardo M Airoldi89,422, Slimane Ait-Si-Ali1376,
\nTakahiko Akematsu2010, Emmanuel T Akporiaye1097, Mohamed Al-Rubeai1394, Guillermo M Albaiceta1294,
\nChris Albanese363, Diego Albani561, Matthew L Albert517, Jesus Aldudo128, Hana Alg€ul1164, Mehrdad Alirezaei1198,
\nIraide Alloza642,888, Alexandru Almasan206, Maylin Almonte-Beceril524, Emad S Alnemri1212, Covadonga Alonso544,
\nNihal Altan-Bonnet848, Dario C Altieri1205, Silvia Alvarez1497, Lydia Alvarez-Erviti1395, Sandro Alves107,
\nGiuseppina Amadoro860, Atsuo Amano930, Consuelo Amantini1554, Santiago Ambrosio1458, Ivano Amelio756,
\nAmal O Amer918, Mohamed Amessou2089, Angelika Amon726, Zhenyi An1538, Frank A Anania291, Stig U Andersen6,
\nUsha P Andley2079, Catherine K Andreadi1690, Nathalie Andrieu-Abadie502, Alberto Anel2027, David K Ann58,
\nShailendra Anoopkumar-Dukie388, Manuela Antonioli832,858, Hiroshi Aoki1791, Nadezda Apostolova2007,
\nSaveria Aquila1500, Katia Aquilano1876, Koichi Araki292, Eli Arama2098, Agustin Aranda456, Jun Araya591,
\nAlexandre Arcaro1472, Esperanza Arias26, Hirokazu Arimoto1225, Aileen R Ariosa1749, Jane L Armstrong1930,
\nThierry Arnould1773, Ivica Arsov2120, Katsuhiko Asanuma675, Valerie Askanas1924, Eric Asselin1867, Ryuichiro Atarashi794,
\nSally S Atherton369, Julie D Atkin713, Laura D Attardi1131, Patrick Auberger1787, Georg Auburger379, Laure Aurelian1727,
\nRiccardo Autelli1992, Laura Avagliano1029,1755, Maria Laura Avantaggiati364, Limor Avrahami1166, Suresh Awale1986,
\nNeelam Azad404, Tiziana Bachetti568, Jonathan M Backer28, Dong-Hun Bae1933, Jae-sung Bae677, Ok-Nam Bae409,
\nSoo Han Bae2117, Eric H Baehrecke1729, Seung-Hoon Baek17, Stephen Baghdiguian1368,
\nAgnieszka Bagniewska-Zadworna2, Hua Bai90, Jie Bai667, Xue-Yuan Bai1133, Yannick Bailly884,
\nKithiganahalli Narayanaswamy Balaji473, Walter Balduini2002, Andrea Ballabio316, Rena Balzan1711, Rajkumar Banerjee239,
\nG abor B anhegyi1052, Haijun Bao2109, Benoit Barbeau1363, Maria D Barrachina2007, Esther Barreiro467, Bonnie Bartel997,
\nAlberto Bartolom e222, Diane C Bassham550, Maria Teresa Bassi1046, Robert C Bast Jr1273, Alakananda Basu1798,
\nMaria Teresa Batista1578, Henri Batoko1336, Maurizio Battino970, Kyle Bauckman2085, Bradley L Baumgarner1909,
\nK Ulrich Bayer1594, Rupert Beale1553, Jean-Fran¸cois Beaulieu1360, George R. Beck Jr48,294, Christoph Becker336,
\nJ David Beckham1595, Pierre-Andr e B edard749, Patrick J Bednarski301, Thomas J Begley1135, Christian Behl1419,
\nChristian Behrends757, Georg MN Behrens406, Kevin E Behrns1627, Eloy Bejarano26, Amine Belaid490,
\nFrancesca Belleudi1041, Giovanni B enard497, Guy Berchem706, Daniele Bergamaschi983, Matteo Bergami1401,
\nBen Berkhout1441, Laura Berliocchi714, Am elie Bernard1749, Monique Bernard1354, Francesca Bernassola1880,
\nAnne Bertolotti791, Amanda S Bess272, S ebastien Besteiro1351, Saverio Bettuzzi1828, Savita Bhalla913,
\nShalmoli Bhattacharyya973, Sujit K Bhutia838, Caroline Biagosch1159, Michele Wolfe Bianchi520,1378,1381,
\nMartine Biard-Piechaczyk210, Viktor Billes298, Claudia Bincoletto1314, Baris Bingol350, Sara W Bird1128, Marc Bitoun1112,
\nIvana Bjedov1258, Craig Blackstone843, Lionel Blanc1183, Guillermo A Blanco1496, Heidi Kiil Blomhoff1812,
\nEmilio Boada-Romero1297, Stefan B€ockler1464, Marianne Boes1423, Kathleen Boesze-Battaglia1835, Lawrence H Boise286,287,
\nAlessandra Bolino2063, Andrea Boman693, Paolo Bonaldo1823, Matteo Bordi897, J€urgen Bosch608, Luis M Botana1308,
\nJoelle Botti1375, German Bou1405, Marina Bouch e1038, Marion Bouchecareilh1331, Marie-Jos ee Boucher1901,
\nMichael E Boulton481, Sebastien G Bouret1926, Patricia Boya133, Micha€el Boyer-Guittaut1345, Peter V Bozhkov1141,
\nNathan Brady374, Vania MM Braga469, Claudio Brancolini1997, Gerhard H Braus353, Jos e M Bravo-San Pedro299,393,508,1374,
\nLisa A Brennan322, Emery H Bresnick2022, Patrick Brest490, Dave Bridges1939, Marie-Agn es Bringer124, Marisa Brini1822,
\nGlauber C Brito1311, Bertha Brodin631, Paul S Brookes1872, Eric J Brown352, Karen Brown1690, Hal E Broxmeyer480,
\nAlain Bruhat486,1339, Patricia Chakur Brum1893, John H Brumell446, Nicola Brunetti-Pierri315,1171,
\nRobert J Bryson-Richardson781, Shilpa Buch1777, Alastair M Buchan1819, Hikmet Budak1022, Dmitry V Bulavin118,505,1789,
\nScott J Bultman1792, Geert Bultynck665, Vladimir Bumbasirevic1470, Yan Burelle1356, Robert E Burke216,217,
\nMargit Burmeister1750, Peter B€utikofer1473, Laura Caberlotto1987, Ken Cadwell896, Monika Cahova112, Dongsheng Cai24,
\nJingjing Cai2099, Qian Cai1018, Sara Calatayud2007, Nadine Camougrand1343, Michelangelo Campanella1700,
\nGrant R Campbell1525, Matthew Campbell1249, Silvia Campello556,1876, Robin Candau1769, Isabella Caniggia1983,
\nLavinia Cantoni560, Lizhi Cao116, Allan B Caplan1656, Michele Caraglia1051, Claudio Cardinali1043, Sandra Morais Cardoso1579, Jennifer S Carew208, Laura A Carleton874, Cathleen R Carlin101, Silvia Carloni2002,
\nSven R Carlsson1267, Didac Carmona-Gutierrez1643, Leticia AM Carneiro312, Oliana Carnevali971, Serena Carra1318,
\nAlice Carrier120, Bernadette Carroll900, Caty Casas1324, Josefina Casas1116, Giuliana Cassinelli324, Perrine Castets1462,
\nSusana Castro-Obregon214, Gabriella Cavallini1841, Isabella Ceccherini568, Francesco Cecconi253,555,1884,
\nArthur I Cederbaum459, Valent ın Ce~na199,1281, Simone Cenci1323,2064, Claudia Cerella444, Davide Cervia1996,
\nSilvia Cetrullo1478, Hassan Chaachouay2028, Han-Jung Chae187, Andrei S Chagin634, Chee-Yin Chai626,628,
\nGopal Chakrabarti1502, Georgios Chamilos1601, Edmond YW Chan1142, Matthew TV Chan181, Dhyan Chandra1003,
\nPallavi Chandra548, Chih-Peng Chang818, Raymond Chuen-Chung Chang1653, Ta Yuan Chang345, John C Chatham1434,
\nSaurabh Chatterjee1910, Santosh Chauhan527, Yongsheng Che62, Michael E Cheetham1263, Rajkumar Cheluvappa1783,
\nChun-Jung Chen1153, Gang Chen598,1676, Guang-Chao Chen9, Guoqiang Chen1078, Hongzhuan Chen1077, Jeff W Chen1514,
\nJian-Kang Chen370,371, Min Chen249, Mingzhou Chen2104, Peiwen Chen1823, Qi Chen1674, Quan Chen172,
\nShang-Der Chen138, Si Chen325, Steve S-L Chen10, Wei Chen2125, Wei-Jung Chen829, Wen Qiang Chen979, Wenli Chen1113,
\nXiangmei Chen1133, Yau-Hung Chen1157, Ye-Guang Chen1250, Yin Chen1447, Yingyu Chen953,955, Yongshun Chen2135,
\nYu-Jen Chen712, Yue-Qin Chen1145, Yujie Chen1208, Zhen Chen339, Zhong Chen2123, Alan Cheng1702,
\nChristopher HK Cheng184, Hua Cheng1728, Heesun Cheong814, Sara Cherry1836, Jason Chesney1703,
\nChun Hei Antonio Cheung817, Eric Chevet1359, Hsiang Cheng Chi140, Sung-Gil Chi656, Fulvio Chiacchiera308,
\nHui-Ling Chiang958, Roberto Chiarelli1826, Mario Chiariello235,567,577, Marcello Chieppa835, Lih-Shen Chin290,
\nMario Chiong1285, Gigi NC Chiu878, Dong-Hyung Cho676, Ssang-Goo Cho650, William C Cho982, Yong-Yeon Cho105,
\nYoung-Seok Cho1064, Augustine MK Choi2095, Eui-Ju Choi656, Eun-Kyoung Choi387,400,685, Jayoung Choi1563,
\nMary E Choi2093, Seung-Il Choi2116, Tsui-Fen Chou412, Salem Chouaib395, Divaker Choubey1574, Vinay Choubey1936,
\nKuan-Chih Chow822, Kamal Chowdhury730, Charleen T Chu1856, Tsung-Hsien Chuang827, Taehoon Chun657,
\nHyewon Chung652, Taijoon Chung978, Yuen-Li Chung1194, Yong-Joon Chwae18, Valentina Cianfanelli254,
\nRoberto Ciarcia1775, Iwona A Ciechomska886, Maria Rosa Ciriolo1876, Mara Cirone1042, Sofie Claerhout1694,
\nMichael J Clague1698, Joan Cl aria1457, Peter GH Clarke1687, Robert Clarke361, Emilio Clementi1045,1398, C edric Cleyrat1781,
\nMiriam Cnop1366, Eliana M Coccia574, Tiziana Cocco1459, Patrice Codogno1375, J€orn Coers271, Ezra EW Cohen1533,
\nDavid Colecchia235,567,577, Luisa Coletto25, N uria S Coll123, Emma Colucci-Guyon516, Sergio Comincini1829,
\nMaria Condello578, Katherine L Cook2073, Graham H Coombs1929, Cynthia D Cooper2076, J Mark Cooper1395,
\nIsabelle Coppens601, Maria Tiziana Corasaniti1387, Marco Corazzari485,1884, Ramon Corbalan1566,
\nElisabeth Corcelle-Termeau251, Mario D Cordero1899, Cristina Corral-Ramos1289, Olga Corti507,1109, Andrea Cossarizza1767,
\nPaola Costelli1993, Safia Costes1518, Susan L Cotman721, Ana Coto-Montes946, Sandra Cottet566,1688, Eduardo Couve1301,
\nLori R Covey1015, L Ashley Cowart762, Jeffery S Cox1536, Fraser P Coxon1427, Carolyn B Coyne1846, Mark S Cragg1919,
\nRolf J Craven1679, Tiziana Crepaldi1995, Jose L Crespo1300, Alfredo Criollo1285, Valeria Crippa558, Maria Teresa Cruz1576,
\nAna Maria Cuervo26, Jose M Cuezva1277, Taixing Cui1907, Pedro R Cutillas987, Mark J Czaja27, Maria F Czyzyk-Krzeska1572,
\nRuben K Dagda2068, Uta Dahmen1404, Chunsun Dai800, Wenjie Dai1187, Yun Dai2059, Kevin N Dalby1940,
\nLuisa Dalla Valle1822, Guillaume Dalmasso1340, Marcello D’Amelio557, Markus Damme188, Arlette Darfeuille-Michaud1340,
\nCatherine Dargemont950, Victor M Darley-Usmar1433, Srinivasan Dasarathy205, Biplab Dasgupta202, Srikanta Dash1254,
\nCrispin R Dass242, Hazel Marie Davey8, Lester M Davids1560, David D avila227, Roger J Davis1731, Ted M Dawson604,
\nValina L Dawson606, Paula Daza1898, Jackie de Belleroche470, Paul de Figueiredo1180,1182,
\nRegina Celia Bressan Queiroz de Figueiredo135, Jos e de la Fuente1023, Luisa De Martino1775,
\nAntonella De Matteis1171, Guido RY De Meyer1443, Angelo De Milito631, Mauro De Santi2002,

Face detection and alignment in unconstrained environment are challenging due to various poses, illuminations, and occlusions. Recent studies show that deep learning approaches can achieve impressive performance on these two tasks. In this letter, we propose a deep cascaded multitask framework that exploits the inherent correlation between detection and alignment to boost up their performance. In particular, our framework leverages a cascaded architecture with three stages of carefully designed deep convolutional networks to predict face and landmark location in a coarse-to-fine manner. In addition, we propose a new online hard sample mining strategy that further improves the performance in practice. Our method achieves superior accuracy over the state-of-the-art techniques on the challenging face detection dataset and benchmark and WIDER FACE benchmarks for face detection, and annotated facial landmarks in the wild benchmark for face alignment, while keeps real-time performance.

In this paper, we propose a simple but effective image prior - dark channel prior to remove haze from a single input image. The dark channel prior is a kind of statistics of the haze-free outdoor images. It is based on a key observation - most local patches in haze-free outdoor images contain some pixels which have very low intensities in at least one color channel. Using this prior with the haze imaging model, we can directly estimate the thickness of the haze and recover a high quality haze-free image. Results on a variety of outdoor haze images demonstrate the power of the proposed prior. Moreover, a high quality depth map can also be obtained as a by-product of haze removal.

Single image haze removal has been a challenging problem due to its ill-posed nature. In this paper, we propose a simple but powerful color attenuation prior for haze removal from a single input hazy image. By creating a linear model for modeling the scene depth of the hazy image under this novel prior and learning the parameters of the model with a supervised learning method, the depth information can be well recovered. With the depth map of the hazy image, we can easily estimate the transmission and restore the scene radiance via the atmospheric scattering model, and thus effectively remove the haze from a single image. Experimental results show that the proposed approach outperforms state-of-the-art haze removal algorithms in terms of both efficiency and the dehazing effect.

A large amount of sequencing data is generated and processed every day with the continuous evolution of sequencing technology and the expansion of sequencing applications. One consequence of such sequencing data explosion is the increasing cost and complexity of data processing. The preprocessing of FASTQ data, which means removing adapter contamination, filtering low-quality reads, and correcting wrongly represented bases, is an indispensable but resource intensive part of sequencing data analysis. Therefore, although a lot of software applications have been developed to solve this problem, bioinformatics scientists and engineers are still pursuing faster, simpler, and more energy-efficient software. Several years ago, the author developed fastp, which is an ultrafast all-in-one FASTQ data preprocessor with many modern features. This software has been approved by many bioinformatics users and has been continuously maintained and updated. Since the first publication on fastp, it has been greatly improved, making it even faster and more powerful. For instance, the duplication evaluation module has been improved, and a new deduplication module has been added. This study aimed to introduce the new features of fastp and demonstrate how it was designed and implemented.

Recent advances in clothes recognition have been driven by the construction of clothes datasets. Existing datasets are limited in the amount of annotations and are difficult to cope with the various challenges in real-world applications. In this work, we introduce DeepFashion1, a large-scale clothes dataset with comprehensive annotations. It contains over 800,000 images, which are richly annotated with massive attributes, clothing landmarks, and correspondence of images taken under different scenarios including store, street snapshot, and consumer. Such rich annotations enable the development of powerful algorithms in clothes recognition and facilitating future researches. To demonstrate the advantages of DeepFashion, we propose a new deep model, namely FashionNet, which learns clothing features by jointly predicting clothing attributes and landmarks. The estimated landmarks are then employed to pool or gate the learned features. It is optimized in an iterative manner. Extensive experiments demonstrate the effectiveness of FashionNet and the usefulness of DeepFashion.

Face detection is one of the most studied topics in the computer vision community. Much of the progresses have been made by the availability of face detection benchmark datasets. We show that there is a gap between current face detection performance and the real world requirements. To facilitate future face detection research, we introduce the WIDER FACE dataset1, which is 10 times larger than existing datasets. The dataset contains rich annotations, including occlusions, poses, event categories, and face bounding boxes. Faces in the proposed dataset are extremely challenging due to large variations in scale, pose and occlusion, as shown in Fig. 1. Furthermore, we show that WIDER FACE dataset is an effective training source for face detection. We benchmark several representative detection systems, providing an overview of state-of-the-art performance and propose a solution to deal with large scale variation. Finally, we discuss common failure cases that worth to be further investigated.

The key challenge of face recognition is to develop effective feature representations for reducing intra-personal variations while enlarging inter-personal differences. In this paper, we show that it can be well solved with deep learning and using both face identification and verification signals as supervision. The Deep IDentification-verification features (DeepID2) are learned with carefully designed deep convolutional networks. The face identification task increases the inter-personal variations by drawing DeepID2 extracted from different identities apart, while the face verification task reduces the intra-personal variations by pulling DeepID2 extracted from the same identity together, both of which are essential to face recognition. The learned DeepID2 features can be well generalized to new identities unseen in the training data. On the challenging LFW dataset, 99.15% face verification accuracy is achieved. Compared with the best deep learning result on LFW, the error rate has been significantly reduced by 67%.

A novel low-cost aluminum–graphite dual-ion battery is reported. The battery shows a reversible capacity of ≈100 mAh g−1 and a capacity retention of 88% after 200 charge–discharge cycles. A packaged aluminum–graphite battery is estimated to deliver an energy density of ≈150 Wh kg−1 at a power density of ≈1200 W kg−1, which is ≈50% higher than most commercial lithium ion batteries. Lithium ion batteries based on cation intercalation have been powering the increasingly mobile society for decades.1 In a conventional lithium ion battery, the intercalation of lithium ions in both cathode (i.e., LiCoO2, LiFePO4) and anode (i.e., graphite, silicon) materials have been thoroughly studied, while the utilization of the anions in the electrolyte has drawn much less attention.2 In fact, the phenomenon of anion intercalate into graphite by chemical or electrochemical means was discovered and proposed as a possible positive electrode for batteries by Rüdorff and Hofmann in 1938.3 However, the anion intercalation was achieved by using high concentration acid solution as electrolyte, this brought serious safety issue that hindered its application.4 In the 1990s, soon after the commercial application of lithium ion battery, Carlin et al. reported dual graphite intercalating molten electrolyte batteries that realized the application of anion intercalated graphite as positive electrode in batteries by using room temperature ionic liquids as electrolyte.5 In the following decades, continuous progresses have been made in anion intercalated graphite based dual carbon batteries, such as investigation of anion intercalation in non-aqueous electrolyte, in situ characterization of the staged anion intercalation process, and systematic study of the intercalation of different anions into graphite.6 However, due to electrolyte decomposition caused by the high positive potential of anion intercalated graphite (≈5 V vs Li/Li+) and exfoliation of graphite layers upon repeated ion/solvent molecule intercalation/deintercalation, reported dual-carbon batteries showed unsatisfied charge–discharge reversibility.7 A key challenge of developing highly reversible dual-graphite battery is to find a suitable electrolyte enabling both Li+ intercalation into graphite negative electrode and anion intercalation into graphite positive electrode simultaneously. Conventional carbonate electrolytes were mainly composed of ethylene carbonate (EC), acyclic carbonate, and lithium salt. EC in the electrolyte is an important component for the formation of solid electrolyte interphase (SEI) and the protection of the negative graphitic electrode.8 Unfortunately, when applied in a dual-graphite battery, the EC molecules in the electrolyte can bind tightly with PF6− anions, and prevent the intercalation of these anions into the interlayer spaces of graphite positive electrodes.9 Recently, with the developments of novel electrolyte formulas, several studies have reported significantly improved reversibility of dual-carbon batteries.10 Read et al. reported a reversible dual-graphite battery with simultaneous accommodation of Li+ and PF6− in graphitic structures enabled by a high voltage electrolyte based on fluorinated solvent and additive.[10] The battery demonstrated a reversible capacity of 60 mAh g−1 and a capacity retention of 62% after 50 cycles at C/7 rate. Rothermel et al. reported a dual-graphite battery based on a mixture of lithium bis-(trifluoromethanesulfonyl)-imide (LiTFSI) and ionic liquid with SEI-forming additive. This electrolyte formula not only enabled stable TFSI− intercalation into the graphite positive electrode, but also allowed highly reversible intercalation of Li+ into the graphite negative electrode.[10] Under an upper cut-off potential of 5.0 V, the full graphite battery presented a capacity of 97 mAh g−1 at a current rate of 10 mA g−1, and 50 mAh g−1 at 500 mA g−1, which shed light on the potential application of dual-ion batteries as an environmentally friendly energy storage technology. Herein, we report a novel aluminum–graphite dual-ion battery (AGDIB) in an ethyl–methyl carbonate (EMC) electrolyte with high reversibility and high energy density. It is the first report on using an aluminum anode in dual-ion battery. The battery shows good reversibility, delivering a capacity of ≈100 mAh g−1 and capacity retention of 88% after 200 charge–discharge cycles at 2 C (1 C corresponding to 100 mA g−1). To the best of our knowledge, performance of the battery is among the best of reported dual-ion batteries. Figure 1a schematically illustrates the initial and charged states of the AGDIB. Upon charging, PF6− anions in the electrolyte intercalate into the graphite cathode, while the Li+ ions in the electrolyte deposit onto the aluminum counter electrode to form an Al–Li alloy. The discharge process is the reverse of the charge process, where both PF6− anions and Li+ ions diffuse back into the electrolyte. The Al counter electrode acts as both the anode and the current collector, which greatly benefits the specific energy density and volumic energy density of the AGDIB.11 Figure 1b shows galvanostatic charge–discharge curves of the AGDIB, exhibiting a typical profile of anion intercalation/deintercalation into/from graphite. The charge curve is mainly composed of three regions between 4.08 and 4.59 V (stage III), 4.59 and 4.63 V (stage II), and 4.63 and 5.0 V (stage I), each region corresponds to an anion intercalation stage of graphite, according to previous reports.[6] A dQ/dV differential curve of the battery is shown in the inset of Figure 1b. Peaks in the profile correspond to electrochemical processes in the AGDIB during charge–discharge. Stage III contains three wide weak peaks, while stage II contains a strong peak and small shoulder peak. No obvious peak in stage I can be observed. Only three peaks are found in the discharge process, due to the diminish or disappear of two peaks during anion deintercalation process, as reported in previous report.[10] The AGDIB shows a working voltage range of 4.8–3.4 V with a middle working voltage of ≈4.2 V at 0.5 C current rate, which is much higher than most of those in commercial lithium ion batteries (≈3.7 V).12 The relative high discharge voltage of the AGDIB enabled a single coin cell to light up two yellow LEDs (nominal voltage of ≈2.5 V) in series (Figure 1b inset). We believe that the good performance of the AGDIB battery is resulted from its specially designed configuration. Firstly, unlike other reported dual-ion batteries, we use aluminum instead of graphite anode. This design eliminates the needs for an addition metallic current collector and lead to considerable weight saving. Another advantage is that we can now eliminate the use of EC, which is commonly used in the electrolyte for protecting the graphite anode. This allows us to use a 100% EMC solvent in the electrolyte which not only solve the problem of binding between EC and PF6−, as reported by Wand and Gao et al.,9, 13 it can also dissolve a much higher concentration of LiPF6 comparing to the commonly used mixed solvents with EC. Seel and Dahn14 have shown that using a high salt concentration can reduce the potential required for anion intercalation into the graphite cathode. Figure S2 in the Supporting Information shows charge–discharge curves in Li | LiPF6 in EMC | Graphite batteries with different LiPF6 concentrations. As the concentration of LiPF6 increases from 1 to 4 m, the anion intercalation potential decreases from 4.45 to 4.34 V (Figure S2b, Supporting Information). This leads to a corresponding increase in specific capacity from 54 to 84 mAh g−1. We also found that performance of the battery depends critical on the metal used in the anode. We tested metal | 4 m LiPF6 in EMC | graphite batteries with different metal counter electrodes (Cu, Fe, Li, Al). Surprisingly, the aluminum | 4 m LiPF6 in EMC | graphite battery shows much higher initial discharge capacity and coulombic efficiency than the other batteries (Figure 2). As both Cu and Fe are inert to form alloy with lithium, the batteries based on Cu and Fe counter electrodes were unable to maintain the lithium ions deposited on them during the charge process. Therefore, only limited discharge capacities were observed in these batteries. On contrary, aluminum is able to form alloy with lithium and it has been studied as a promising anode material for lithium ion battery.15 During charge process, lithium ions in the electrolyte obtain electrons on the surface of aluminum counter electrode and form stable aluminum–lithium alloy. This alloy enabled the controlled release of the lithium ions in the discharge process. Although the Al | 4 m LiPF6 in EMC | Graphite battery exhibits impressive initial discharge capacity, the cycle stability of this battery is poor, resulting from the pulverization caused by the volume expansion of aluminum during the alloying process.[15] To improve the cycle stability, we tried to add SEI formation additive into the electrolyte to protect the aluminum counter electrode from pulverization. After screening of different electrolyte additives, we found that a small amount of vinylene carbonate (VC) is very effective in improving the cycle stability of the aluminum–graphite battery. Figure S3 in the Supporting Information shows the charge–discharge curves of the formation process of the AGDIB with different amount of VC in the electrolyte. With the presence of VC, the charge curves show an extra plateau at about 4.37 V, which corresponds to the decomposition of VC and the formation of SEI layer.[8] Meanwhile, the amount of VC was found to be critical for improving the cycle stability of the AGDIB. When the amount of VC is low (i.e., 1 wt%), the battery showed fast capacity decay (Figure S4a, Supporting Information). On the other hand, high amount of VC (i.e., 5 wt%) will lead to relatively lower coulombic efficiency during cycling and a battery failure was tricked after several charge–discharge cycles (Figure S4b, Supporting Information). We found that batteries with 2 wt% VC in the electrolyte showed the best stability. Figure 3 shows the charge–discharge cycle test results of aluminum–graphite battery with commercial carbonate electrolyte (1 m LiPF6 in EC/EMC/DMC), 4 m LiPF6 in EMC, and 4 m LiPF6 in EMC + 2 wt% VC. With the presence of VC additive, the AGDIB exhibits a reversible discharge capacity of 105 mAh g−1 (the seventh cycle) and a capacity retention of 96% after 50 cycles at 0.5 C current rate. For comparison, the capacity of the batteries without VC additive dramatically decayed within the first 20 charge–discharge cycles. The battery with 2 wt% VC additive was then cycled at various charge–discharge rates ranging from 0.5 to 5 C over a potential window of 3.0–5.0 V. Typical galvanostatic profiles of the battery are shown in Figure 4a. All the curves show the typical three-stage charge–discharge profile of anion intercalation in graphite, as discussed before. The gradually increased charge–discharge plateau separation implies small electrode polarization at low current rate and relatively large polarization at high current rates. Figure 4b further shows the charge–discharge capacities and corresponding coulombic efficiencies of the AGDIB during the rate capacity tests. At the rates of 0.5, 1, 2, 3, and 5 C, the battery show discharge capacities of 105, 104, 100, 93, and 79 mAh g−1, respectively. The battery can regain the high capacity when the current rate was set back to lower value gradually, demonstrating its high reversibility. The gradually increasing coulombic efficiencies (67%–83%) during the first few cycles at 0.5 C, was attributed to the formation of protective SEI layer on the surface of the electrodes. After the first 10 cycles at 0.5 C, the battery shows stable coulombic efficiencies at each of the following current rates, with values of 91%, 92%, 96%, and 98% at 1, 2, 3, and 5 C, respectively. Figure 4c illustrates the long term cycling performance of the AGDIB at a current rate of 2 C. Compared with batteries based on anion intercalated graphite reported in literature,[7, 10, 16] this AGDIB cell shows much improved cyclability. Notably, the discharge capacity varies from 104 to 92 mAh g−1 during the 200 charge–discharge cycles, corresponding to capacity retention of 88% and only 0.06% capacity loss per cycle. The electrochemical performance of this AGDIB is among the best of reported dual-ion batteries (Table S3). The AGDIB reported here was composed of only environmentally friendly low cost materials (i.e., aluminum and graphite) as electrode materials, and conventional lithium salt and carbonate solvent as electrolyte. Compared with conventional secondary battery technologies (mainly lithium ion batteries), it shows an obvious advantage in production cost. Furthermore, as the Al counter electrode in the AGDIB acts as both the anode and the current collector, the dead load and dead volume of this battery could be significantly reduced, which result in a battery with both high specific energy density and high volume energy density. We roughly estimated the specific energy density and power density of the AGDIB based on the tested results and the mass composition of conventional packaged batteries (detail of the calculation is available in supporting materials). The calculation results (Table S1) show that the AGDIB can deliver a specific energy density of ≈222 Wh kg−1 at a power density of 132 W kg−1, and ≈150 Wh kg−1 at 1200 W kg−1. Figure 4d shows a comparison of the AGDIB with several main-stream energy storage technologies. Apparently, comparing with commercial lithium ion battery (≈200 Wh kg−1 at 50 W kg−1, and ≈100 Wh kg−1 at 1000 W kg−1) and electrochemical capacitor (≈5 Wh kg−1 at 5000 W kg−1), the AGDIB shows significantly improved performances.17 To understand the charge–discharge mechanism of the AGDIB, further characterizations of the electrodes were carried out. The anion intercalation in graphite positive electrode is a reported phenomenon. The intercalation of PF6− (size of 4.36 Å)18 into graphite sheets (distance of 3.36 Å) is accompanied by significant inter space expansion and gradual exfoliation of the graphite sheets (Figure S5, Supporting Information), which is responsible for the slow capacity degradation during cycling. The electrochemical process on the Al counter electrode is of great interests. The charge–discharge curves of a Li–Al half cell is shown in Figure S6 in the Supporting Information. It shows that the Al–Li alloying process on the aluminum electrode exhibited a flat plateau at about 0.22 V versus Li/Li+, and 0.52 V versus Li/Li+ during dealloying process. Figure 5a,b shows photographs of the Al electrodes before and after charged in the AGDIB. A rough layer on the surface of the charged Al electrode could be observed by naked eyes. Scanning electron microscopic (SEM) image shows that the rough layer possess nanoporous microstructures (Figure 5d), which was likely to be caused by the formation of SEI layer and the Al–Li alloying process. Energy dispersive X-ray spectroscopy (EDS) mapping (Figure S7 and Table S2, Supporting Information) on the surface of the charged Al electrode show the existence of F, C, and O, which are common composing elements of SEI layers.19 This SEI layer could protect the Al electrode from destruction during Li–Al alloying and dealloying process. For comparison, much less elements component of SEI layer were found (Figure S8, Supporting Information) on the surface of Al electrode without the presence of VC. Without the protection of SEI layer, pulverization and cracks were found on the surface of the Al electrode (Figure S9, Supporting Information).The existence of Al–Li alloy on the charged Al electrode was verified by the XRD patterns (Figure 5e), where diffraction peaks of both Al (JCPDS Card No. 65-2869) and AlLi (JCPDS Card No. 65-3017) can be clearly observed. According to the characterization results, electrochemical reactions in the AGDIB are proposed as follows. The Al–Li alloying process has been studied as a potential anode of lithium ion battery in previous works.15, 20 The theoretical capacity of aluminum could reach 2235 mAh g−1 in the form of Li9Al4. However, the lithiation of aluminum faces a ≈100% volume expansion, which would tear up the anode SEI layer and cause the pulverization of the active material. Here in the AGDIB, the lithiation of Al electrode was stabilized for the following two reasons. Firstly, compared with the deep lithiation in Li9Al4, the volume expansion caused by the shallow lithiation (AlLi) in the present AGDIB is smaller, which could help reduce the stress in the Al electrode caused by lithiation. Secondly, VC induced SEI layer on the surface of Al electrode could protect the Al electrode from destruction caused by the Li–Al alloying process. Detailed mechanism of this AGDIB battery, in particularly the role of VC are still under further investigation. In summary, we have developed a novel AGDIB composed of only environmentally friendly low-cost materials (i.e., aluminum as counter electrode, graphite as positive electrode), and a specially designed carbonate electrolyte. As aluminum acted as both the negative current collector and the negative active material, the AGDIB shows significantly reduced dead load and dead volume. The AGDIB delivers a reversible capacity of 104 mAh g−1 (based on the mass of graphite) at 2 C current rate, and a capacity retention of 88% after 200 cycles. According to the composition of conventional packaged battery, a packaged AGDIB cell is estimated to delivery an energy density of ≈220 Wh kg−1 at a power density of ≈130 W kg−1, and ≈150 Wh kg−1 at ≈1200 W kg−1, which are significantly higher than most commercial lithium ion batteries, indicating its potential to be a low-cost power source with both high energy density and high power density. Aluminum foil (thickness of 15 μm) purchased from Shenzhen Kejingstar was used as the counter electrode. The graphite positive electrode composes of 80 wt% of natural graphite (d002 spacing of 3.363 Å, SEM image shown in Figure S1, Supporting Information), 10 wt% of conductive carbon black, and 10 wt% of polyvinylidene fluoride (PVDF) as binder.N-methyl-2-pyrrolidone (NMP) was added to the above mixture and grounded to form a uniform slurry. Then the slurry was pasted onto aluminum foil and dried in vacuum in an 80 °C oven. The mass loading of the graphite positive electrode was 1.5 mg cm−2.The components of the electrolyte, namely EMC, VC, and LiPF6 were used as purchased from Dodochem. The commercial electrolyte, 1 m LiPF6 EC–EMC–DMC (1:1:1), was kindly provided by Capchem. Powder X-ray diffraction (XRD) analysis was carried out on a Rigaku D/Max-2500 diffractometer using Cu Kα radiation with a scan rate of 8° min−1, operating at 40 kV and 30 mA. XRD samples of the charged electrodes were prepared in glove box, washed with EMC, and coated with parafilm to protect them from been oxidized by air. SEM images were collected on a HITACHI S-4800 field emission scanning electron microscope. Electrochemical tests were performed with CR2032 coin-type cells. The graphite positive electrode was countered with different metal electrodes (Li, Al, Fe, Cu). Lab-made electrolyte or the purchased commercial electrolyte were used as the electrolyte and glass fiber was used as separator. The preparation of electrolyte and fabrication of battery were conducted in an argon filled glove box (Etelux Lab2000). Cyclic voltammetry (CV) tests were done on an AUTOLAB PGSTAT302N electrochemical station. Galvanostatic charge–discharge tests were carried out on LAND CT2011A battery test system at room temperature. This project was financially supported by the National Natural Science Foundation of China (Nos. 51272217, 51302238), Science and Technology Planning Project of Guangdong Province (Nos. 2014A010105032, 2014A010106016), Guangdong Innovative and Entrepreneurial Research Team Program (No. 2013C090), Shenzhen Municipality Project (No. JCYJ20140417113430618, JSGG20150602143328010), and Scientific Equipment Project of Chinese Academy of Sciences (yz201440) As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Y Artificial intelligence (AI) coupled with promising machine learning (ML) techniques well known from computer science is broadly affecting many aspects of various fields including science and technology, industry, and even our day-to-day life. The ML techniques have been developed to analyze high-throughput data with a view to obtaining useful insights, categorizing, predicting, and making evidence-based decisions in novel ways, which will promote the growth of novel applications and fuel the sustainable booming of AI. This paper undertakes a comprehensive survey on the development and application of AI in different aspects of fundamental sciences, including information science, mathematics, medical science, materials science, geoscience, life science, physics, and chemistry. The challenges that each discipline of science meets, and the potentials of AI techniques to handle these challenges, are discussed in detail. Moreover, we shed light on new research trends entailing the integration of AI into each scientific discipline. The aim of this paper is to provide a broad research guideline on fundamental sciences with potential infusion of AI, to help motivate researchers to deeply understand the state-of-the-art applications of AI-based fundamental sciences, and thereby to help promote the continuous development of these fundamental sciences.

This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus on proposed solutions and results. A new DIVerse 2K resolution image dataset (DIV2K) was employed. The challenge had 6 competitions divided into 2 tracks with 3 magnification factors each. Track 1 employed the standard bicubic downscaling setup, while Track 2 had unknown downscaling operators (blur kernel and decimation) but learnable through low and high res train images. Each competition had ∽100 registered participants and 20 teams competed in the final testing phase. They gauge the state-of-the-art in single image super-resolution.

Though many attempts have been made in blind super-resolution to restore low-resolution images with unknown and complex degradations, they are still far from addressing general real-world degraded images. In this work, we extend the powerful ESRGAN to a practical restoration application (namely, Real-ESRGAN), which is trained with pure synthetic data. Specifically, a high-order degradation modeling process is introduced to better simulate complex real-world degradations. We also consider the common ringing and overshoot artifacts in the synthesis process. In addition, we employ a U-Net discriminator with spectral normalization to increase discriminator capability and stabilize the training dynamics. Extensive comparisons have shown its superior visual performance than prior works on various real datasets. We also provide efficient implementations to synthesize training pairs on the fly.

Metal-halide perovskites have rapidly emerged as one of the most promising materials of the 21st century, with many exciting properties and great potential for a broad range of applications, from photovoltaics to optoelectronics and photocatalysis. The ease with which metal-halide perovskites can be synthesized in the form of brightly luminescent colloidal nanocrystals, as well as their tunable and intriguing optical and electronic properties, has attracted researchers from different disciplines of science and technology. In the last few years, there has been a significant progress in the shape-controlled synthesis of perovskite nanocrystals and understanding of their properties and applications. In this comprehensive review, researchers having expertise in different fields (chemistry, physics, and device engineering) of metal-halide perovskite nanocrystals have joined together to provide a state of the art overview and future prospects of metal-halide perovskite nanocrystal research.

Materials exhibiting high energy/power density are currently needed to meet the growing demand of portable electronics, electric vehicles and large-scale energy storage devices. The highest energy densities are achieved for fuel cells, batteries, and supercapacitors, but conventional dielectric capacitors are receiving increased attention for pulsed power applications due to their high power density and their fast charge-discharge speed. The key to high energy density in dielectric capacitors is a large maximum but small remanent (zero in the case of linear dielectrics) polarization and a high electric breakdown strength. Polymer dielectric capacitors offer high power/energy density for applications at room temperature, but above 100 °C they are unreliable and suffer from dielectric breakdown. For high-temperature applications, therefore, dielectric ceramics are the only feasible alternative. Lead-based ceramics such as La-doped lead zirconate titanate exhibit good energy storage properties, but their toxicity raises concern over their use in consumer applications, where capacitors are exclusively lead free. Lead-free compositions with superior power density are thus required. In this paper, we introduce the fundamental principles of energy storage in dielectrics. We discuss key factors to improve energy storage properties such as the control of local structure, phase assemblage, dielectric layer thickness, microstructure, conductivity, and electrical homogeneity through the choice of base systems, dopants, and alloying additions, followed by a comprehensive review of the state-of-the-art. Finally, we comment on the future requirements for new materials in high power/energy density capacitor applications.

Humans rely increasingly on sensors to address grand challenges and to improve quality of life in the era of digitalization and big data. For ubiquitous sensing, flexible sensors are developed to overcome the limitations of conventional rigid counterparts. Despite rapid advancement in bench-side research over the last decade, the market adoption of flexible sensors remains limited. To ease and to expedite their deployment, here, we identify bottlenecks hindering the maturation of flexible sensors and propose promising solutions. We first analyze challenges in achieving satisfactory sensing performance for real-world applications and then summarize issues in compatible sensor-biology interfaces, followed by brief discussions on powering and connecting sensor networks. Issues en route to commercialization and for sustainable growth of the sector are also analyzed, highlighting environmental concerns and emphasizing nontechnical issues such as business, regulatory, and ethical considerations. Additionally, we look at future intelligent flexible sensors. In proposing a comprehensive roadmap, we hope to steer research efforts towards common goals and to guide coordinated development strategies from disparate communities. Through such collaborative efforts, scientific breakthroughs can be made sooner and capitalized for the betterment of humanity.

Visual features are of vital importance for human action understanding in videos. This paper presents a new video representation, called trajectory-pooled deep-convolutional descriptor (TDD), which shares the merits of both hand-crafted features [31] and deep-learned features [24]. Specifically, we utilize deep architectures to learn discriminative convolutional feature maps, and conduct trajectory-constrained pooling to aggregate these convolutional features into effective descriptors. To enhance the robustness of TDDs, we design two normalization methods to transform convolutional feature maps, namely spatiotemporal normalization and channel normalization. The advantages of our features come from (i) TDDs are automatically learned and contain high discriminative capacity compared with those hand-crafted features; (ii) TDDs take account of the intrinsic characteristics of temporal dimension and introduce the strategies of trajectory-constrained sampling and pooling for aggregating deep-learned features. We conduct experiments on two challenging datasets: HMD-B51 and UCF101. Experimental results show that TDDs outperform previous hand-crafted features [31] and deep-learned features [24]. Our method also achieves superior performance to the state of the art on these datasets.

Video restoration tasks, including super-resolution, deblurring, etc, are drawing increasing attention in the computer vision community. A challenging benchmark named REDS is released in the NTIRE19 Challenge. This new benchmark challenges existing methods from two aspects: (1) how to align multiple frames given large motions, and (2) how to effectively fuse different frames with diverse motion and blur. In this work, we propose a novel Video Restoration framework with Enhanced Deformable convolutions, termed EDVR, to address these challenges. First, to handle large motions, we devise a Pyramid, Cascading and Deformable (PCD) alignment module, in which frame alignment is done at the feature level using deformable convolutions in a coarse-to-fine manner. Second, we propose a Temporal and Spatial Attention (TSA) fusion module, in which attention is applied both temporally and spatially, so as to emphasize important features for subsequent restoration. Thanks to these modules, our EDVR wins the champions and outperforms the second place by a large margin in all four tracks in the NTIRE19 video restoration and enhancement challenges. EDVR also demonstrates superior performance to state-of-the-art published methods on video super-resolution and deblurring. The code is available at https://github.com/xinntao/EDVR.

International challenges have become the de facto standard for comparative assessment of image analysis algorithms. Although segmentation is the most widely investigated medical image processing task, the various challenges have been organized to focus only on specific clinical tasks. We organized the Medical Segmentation Decathlon (MSD)-a biomedical image analysis challenge, in which algorithms compete in a multitude of both tasks and modalities to investigate the hypothesis that a method capable of performing well on multiple tasks will generalize well to a previously unseen task and potentially outperform a custom-designed solution. MSD results confirmed this hypothesis, moreover, MSD winner continued generalizing well to a wide range of other clinical problems for the next two years. Three main conclusions can be drawn from this study: (1) state-of-the-art image segmentation algorithms generalize well when retrained on unseen tasks; (2) consistent algorithmic performance across multiple tasks is a strong surrogate of algorithmic generalizability; (3) the training of accurate AI segmentation models is now commoditized to scientists that are not versed in AI model training.