Search

Your search keyword '"Qin, Danfeng"' showing total 114 results
Start Over Author "Qin, Danfeng"
114 results on '"Qin, Danfeng"'

1. MobileNetV4 -- Universal Models for the Mobile Ecosystem

Author

Qin, Danfeng, Leichner, Chas, Delakis, Manolis, Fornoni, Marco, Luo, Shixin, Yang, Fan, Wang, Weijun, Banbury, Colby, Ye, Chengxi, Akin, Berkin, Aggarwal, Vaibhav, Zhu, Tenghui, Moro, Daniele, and Howard, Andrew

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

We present the latest generation of MobileNets, known as MobileNetV4 (MNv4), featuring universally efficient architecture designs for mobile devices. At its core, we introduce the Universal Inverted Bottleneck (UIB) search block, a unified and flexible structure that merges Inverted Bottleneck (IB), ConvNext, Feed Forward Network (FFN), and a novel Extra Depthwise (ExtraDW) variant. Alongside UIB, we present Mobile MQA, an attention block tailored for mobile accelerators, delivering a significant 39% speedup. An optimized neural architecture search (NAS) recipe is also introduced which improves MNv4 search effectiveness. The integration of UIB, Mobile MQA and the refined NAS recipe results in a new suite of MNv4 models that are mostly Pareto optimal across mobile CPUs, DSPs, GPUs, as well as specialized accelerators like Apple Neural Engine and Google Pixel EdgeTPU - a characteristic not found in any other models tested. Finally, to further boost accuracy, we introduce a novel distillation technique. Enhanced by this technique, our MNv4-Hybrid-Large model delivers 87% ImageNet-1K accuracy, with a Pixel 8 EdgeTPU runtime of just 3.8ms.

Published
2024

2. MobileNetV4: Universal Models for the Mobile Ecosystem

Author

Qin, Danfeng, Leichner, Chas, Delakis, Manolis, Fornoni, Marco, Luo, Shixin, Yang, Fan, Wang, Weijun, Banbury, Colby, Ye, Chengxi, Akin, Berkin, Aggarwal, Vaibhav, Zhu, Tenghui, Moro, Daniele, Howard, Andrew, Goos, Gerhard, Series Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Leonardis, Aleš, editor, Ricci, Elisa, editor, Roth, Stefan, editor, Russakovsky, Olga, editor, Sattler, Torsten, editor, and Varol, Gül, editor

Published
2025
Full Text
View/download PDF

9. Improving Object Detection with Selective Self-supervised Self-training

Author

Li, Yandong, Huang, Di, Qin, Danfeng, Wang, Liqiang, and Gong, Boqing

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

We study how to leverage Web images to augment human-curated object detection datasets. Our approach is two-pronged. On the one hand, we retrieve Web images by image-to-image search, which incurs less domain shift from the curated data than other search methods. The Web images are diverse, supplying a wide variety of object poses, appearances, their interactions with the context, etc. On the other hand, we propose a novel learning method motivated by two parallel lines of work that explore unlabeled data for image classification: self-training and self-supervised learning. They fail to improve object detectors in their vanilla forms due to the domain gap between the Web images and curated datasets. To tackle this challenge, we propose a selective net to rectify the supervision signals in Web images. It not only identifies positive bounding boxes but also creates a safe zone for mining hard negative boxes. We report state-of-the-art results on detecting backpacks and chairs from everyday scenes, along with other challenging object classes., Comment: Accepted to ECCV 2020

Published
2020

11. Cap2Det: Learning to Amplify Weak Caption Supervision for Object Detection

Author

Ye, Keren, Zhang, Mingda, Kovashka, Adriana, Li, Wei, Qin, Danfeng, and Berent, Jesse

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Learning to localize and name object instances is a fundamental problem in vision, but state-of-the-art approaches rely on expensive bounding box supervision. While weakly supervised detection (WSOD) methods relax the need for boxes to that of image-level annotations, even cheaper supervision is naturally available in the form of unstructured textual descriptions that users may freely provide when uploading image content. However, straightforward approaches to using such data for WSOD wastefully discard captions that do not exactly match object names. Instead, we show how to squeeze the most information out of these captions by training a text-only classifier that generalizes beyond dataset boundaries. Our discovery provides an opportunity for learning detection models from noisy but more abundant and freely-available caption data. We also validate our model on three classic object detection benchmarks and achieve state-of-the-art WSOD performance. Our code is available at https://github.com/yekeren/Cap2Det., Comment: To appear in ICCV 2019

Published
2019

13. Learning to discover and localize visual objects with open vocabulary

Author

Ye, Keren, Zhang, Mingda, Li, Wei, Qin, Danfeng, Kovashka, Adriana, and Berent, Jesse

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

To alleviate the cost of obtaining accurate bounding boxes for training today's state-of-the-art object detection models, recent weakly supervised detection work has proposed techniques to learn from image-level labels. However, requiring discrete image-level labels is both restrictive and suboptimal. Real-world "supervision" usually consists of more unstructured text, such as captions. In this work we learn association maps between images and captions. We then use a novel objectness criterion to rank the resulting candidate boxes, such that high-ranking boxes have strong gradients along all edges. Thus, we can detect objects beyond a fixed object category vocabulary, if those objects are frequent and distinctive enough. We show that our objectness criterion improves the proposed bounding boxes in relation to prior weakly supervised detection methods. Further, we show encouraging results on object detection from image-level captions only.

Published
2018

15. Sensitive Electrochemical Determination of Quercetin in Fruit Juice Using an Aluminum Hydroxide Oxide–Titanium Carbide MXene Composite.

Author

Chen, Yidan and Qin, Danfeng

Subjects
*HYBRID materials, *TITANIUM carbide, *CYCLIC voltammetry, *ALUMINUM hydroxide, *ELECTROCHEMICAL sensors
Abstract

AbstractA hybrid material composed of aluminum hydroxide oxide (AlOOH) and titanium carbide (Ti3C2) nanosheets was synthesized and used for electrochemical sensing of quercetin. The Ti3C2 nanosheets were synthesized using a selective etching technique, followed by the hydrothermal growth of AlOOH on the nanosheet surfaces. Cyclic voltammetry and differential pulse voltammetry were performed to evaluate the electrochemical properties of the AlOOH@Ti3C2 nanocomposites. Under optimal conditions, the AlOOH@Ti3C2 modified electrode exhibited wide linearity and high sensitivity. The practicality of the sensor was verified through an interference study and real sample analysis. Additionally, the sensing platform exhibited outstanding operational and storage stability. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

18. Analysis of the miR482 Gene Family in Plants.

Author

Kuang, Wei, Qin, Danfeng, Huang, Ying, Liu, Yihua, Cao, Xue, and Xu, Meng

Subjects
*COMPETITIVE endogenous RNA, *GENE families, *PLANT genes, *URACIL, *MICRORNA
Abstract

MicroRNA482 (miR482) is a conserved microRNA family in plants, playing critical regulatory roles in different biological activities. Though the members of the miR482 gene family have been identified in plants, a systematic study has not been reported yet. In the present research, 140 mature sequences generated by 106 precursors were used for molecular characterization, phylogenetic analysis, and target gene prediction, and the competing endogenous RNA (ceRNA) network mediated by miR482 was summarized. The length of mature sequences ranged from 17 nt to 25 nt, with 22 nt being the most abundant, and the start and end of the mature sequences had a preference for uracil (U). By sequence multiplex comparison, it was found that the mature sequences of 5p were clustered into one group, and others were clustered into the other group. Phylogenetic analysis revealed that the 140 mature sequences were categorized into six groups. Meanwhile, all the precursor sequences formed a stable hairpin structure, and the 106 precursors were divided into five groups. However, the expression of miR482 varied significantly between different species and tissues. In total, 149 target genes were predicted and their functions focused on single-organism process, cellular process, and cell and cell part. The ceRNA network of miR482 in tomato, cotton, and peanut was summarized based on related publications. In conclusion, this research will provide a foundation for further understanding of the miR482 gene family. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

22. Highly Sensitive Electrochemical Determination of Luteolin Using Self-Assembled Titanium Carbide/Single-Walled Carbon Nanohorns (SWCNHs) and Differential Pulsed Stripping Voltammetry (DPV).

Author

Feng, Zhiyang, Zhang, Liwen, Huang, Wenyi, Li, Lijun, Qin, Danfeng, and Cheng, Hao

Subjects
CARBON nanohorns, VOLTAMMETRY, LUTEOLIN, CARBON electrodes, ELECTROCHEMICAL sensors, X-ray photoelectron spectroscopy
Abstract

A novel electrochemical sensor based upon a Ti3C2/single-walled carbon nanohorn (SWCNH) composite was constructed using self-assembly for the determination of luteolin. The doping of SWCNHs avoids the restacking of Ti3C2 nanosheets, thus significantly improving the electrochemical properties of Ti3C2/SWCNHs. The Ti3C2/SWCNH nanocomposite was characterized by scanning electron microscopy (SEM), transmission electron microscopy (TEM), and X-ray photoelectron spectroscopy (XPS). The performance of the sensor was investigated by differential pulse stripping voltammetry (DPV). The Ti3C2/SWCNH-modified glassy carbon electrode exhibits enhanced electrocatalytic activity for luteolin. The composite electrode demonstrates excellent conductivity and high catalysis due to the synergistic properties of Ti3C2 with SWCNHs, which provides high selectivity, good reproducibility, and a low detection limit. Under the optimal conditions, the peak current of the sensor increases linearly with the luteolin concentration from 0.5 to 20 μM, with a detection limit of 0.0086 μM. The sensor was employed for the determination of luteolin in chrysanthemum tea with recoveries between 101.6% and 104.9%. Therefore, the electrochemical sensor based upon the Ti3C2/SWCNH composite is expected to have applications in food and pharmaceutical analysis. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

24. Design and analysis of a 2D grapheneplus (G+)-based gas sensor for the detection of multiple organic gases.

Author

Qin, Danfeng, Chen, Tong, Xie, Luzhen, Yang, Ning, Luo, Cheng, and Zhou, Guanghui

Abstract

A new member of the 2D carbon family, grapheneplus (G+), has demonstrated excellent properties, such as Dirac cones and high surface area. In this study, the electronic transport properties of G+, NG+, and BG+ monolayers in which the NG+/BG+ can be obtained by replacing the center sp3 hybrid carbon atoms of the G+ with N/B atoms, were studied and compared using density functional theory and the non-equilibrium Green's function method. The results revealed that G+ is a semi-metal with two Dirac cones, which becomes metallic upon doping with N or B atoms. Based on the electronic structures, the conductivities of the 2D G+, NG+ and BG+-based nanodevices were analyzed deeply. It was found that the currents of all the designed devices increased with increasing the applied bias voltage, showing obvious quasi-linear current–voltage characteristics. IG+ was significantly higher than ING+ and IBG+ at the same bias voltage, and IG+ was almost twice IBG+, indicating that the electron mobility of G+ can be controlled by B/N doping. Additionally, the gas sensitivities of G+, NG+, and BG+-based gas sensors in detecting C2H4, CH2O, CH4O, and CH4 organic gases were studied. All the considered sensors can chemically adsorb C2H4 and CH2O, but there were only weak van der Waals interactions with CH4O and CH4. For chemical adsorption, the gas sensitivities of these sensors were considerably high and steady, and the sensitivity of NG+ to adsorb C2H4 and CH2O was greater as compared to G+ and BG+ at higher bias voltages. Interestingly, the maximum sensitivity difference for BG+ toward C2H4 and CH2O was 17%, which is better as compared to G+ and NG+. The high sensitivity and different response signals of these sensors were analyzed by transmission spectra and scattering state separation at the Fermi level. Gas sensors based on G+ monolayers can effectively detect organic gases such as C2H4 and CH2O, triggering their broad potential application prospects in the field of gas sensing. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

34. Learning to Overcome Noise in Weak Caption Supervision for Object Detection

Author

Unal, Mesut Erhan, Ye, Keren, Zhang, Mingda, Thomas, Christopher, Kovashka, Adriana, Li, Wei, Qin, Danfeng, and Berent, Jesse

Abstract

We propose the first mechanism to train object detection models from weak supervision in the form of captions at the image level. Language-based supervision for detection is appealing and inexpensive: many blogs with images and descriptive text written by human users exist. However, there is significant noise in this supervision: captions do not mention all objects that are shown, and may mention extraneous concepts. We first propose a technique to determine which image-caption pairs provide suitable signal for supervision. We further propose several complementary mechanisms to extract image-level pseudo labels for training from the caption. Finally, we train an iterative weakly-supervised object detection model from these image-level pseudo labels. We use captions from four datasets (COCO, Flickr30K, MIRFlickr1M, and Conceptual Captions) whose level of noise varies. We evaluate our approach on two object detection datasets. Weighting the labels extracted from different captions provides a boost over treating all captions equally. Further, our primary proposed technique for inferring pseudo labels for training at the image level, outperforms alternative techniques under a wide variety of settings. Both techniques generalize to datasets beyond the one they were trained on.

Published
2023
Full Text
View/download PDF

39. N-Doped Hollow Porous Carbon Spheres/Bismuth Hybrid Film Modified Electrodes for Sensitive Voltammetric Determination of Trace Cadmium

Author

Qin, Danfeng, Wang, Le, Gao, Sanshuang, Wang, Ying, Mamat, Xamxikamar, Li, Yongtao, Wågberg, Thomas, Cheng, Hao, Hu, Guangzhi, Qin, Danfeng, Wang, Le, Gao, Sanshuang, Wang, Ying, Mamat, Xamxikamar, Li, Yongtao, Wågberg, Thomas, Cheng, Hao, and Hu, Guangzhi

Abstract

In this work, N-doped hollow porous carbon spheres (N-HPCSs) were synthesized by silicon dioxide template-assisted polybenzoxazine (PB) coating strategy. The prepared N-HPCSs have a smooth hollow ball structure surrounded by a well-defined porous shell. Combining with in-situ plating of Bi film, the N-HPCSs were further fabricated a sensitive electrochemical platform for determination trace levels of Cd(II) by differential pulse anodic stripping voltammetry (DPASV). Under the optimized conditions, the Bi-N-HPCSs based sensor displays a linear response to Cd(II) over the range of 0.5gL(-1) to 150gL(-1). Meanwhile, the limit of detection (LOD, S/N=3) is estimated to be around 0.16gL(-1) for Cd(II), which is 31 times lower than the safety values set by United States Environmental Protection Agency (EPA) for the drinking water. Moreover, the proposed method was successfully applied to detection of Cd(II) in tap water and lake water, and the analytical results of the presented method are agreed well with inductively coupled plasma-mass spectrometry (ICP-MS) data. Due to the excellent analytical performance, the fabricated electrode is promised for future development in monitoring of cadmium pollution in the environment.

Published
2018
Full Text
View/download PDF

40. An Electrochemical Sensor Based on Green gamma-AlOOH-Carbonated Bacterial Cellulose Hybrids for Simultaneous Determination Trace Levels of Cd(II) and Pb(II) in Drinking Water

Author

Qin, Danfeng, Hu, Xun, Dong, Yemin, Mamat, Xamxikamar, Li, Yongtao, Wågberg, Thomas, Hu, Guangzhi, Qin, Danfeng, Hu, Xun, Dong, Yemin, Mamat, Xamxikamar, Li, Yongtao, Wågberg, Thomas, and Hu, Guangzhi

Abstract

An eco-friendly gamma-AlOOH-carbonated bacterial cellulose (gamma-AlOOH-CBC) hybrids material was fabricated by simple pyrolysis and hydrothermal treatments. The obtained hybrids possess an intrinsic 3D nanofibrous structure decorated with chaff-like gamma-AlOOH particles. Owing to the good adsorption property and conductive, gamma-AlOOH-CBC hybrids were used to modified the glass carbon electrode (GCE) for simultaneous determination of Cd(II) and Pb(II) in aqueous samples by differential pulse anodic stripping voltammetry (DPASV) method. Various parameters affected Cd(II) and Pb(II) measurement were optimized. Under the optimal conditions, the limit of detection (S/N = 3) of the gamma-AlOOH-CBC modified electrode was evaluated to be 0.17 mu g.L-1 for Cd(II) and 0.10 mu g.L-1 for Pb(II) with the linear range of the calibration curves ranged 0.5-250 mu g.L-1 for Cd(II) and Pb(II). Furthermore, the developed electrode was also successfully utilized for monitoring trace Cd(II) and Pb(II) in drinking water samples with satisfactory results.

Published
2018
Full Text
View/download PDF

41. A Novel Electrochemical Sensor for Chloramphenicol Based on MXenes and Carbon Nanofiber from Bacterial Cellulose

Author

Guo, Hangyu, Xu, Feng, Li, Lijun, and Qin, Danfeng

Abstract

In this study, three-dimensional porous MXene/carbon nanofiber (CNF) nanocomposites were prepared by assembling environmentally friendly and inexpensive bacterial cellulose (BC) gel sheets as a carbon source with novel two-dimensional MXenes nanoplate clusters and pyrolyzing the composite. The structure, morphology, and electrochemical properties of MXene/CNF was characterized using scanning electron microscopy, transmission electron microscopy, X-ray diffraction, X-ray photoelectron spectroscopy, and electrochemical techniques. The experimental results revealed that the MXene/CNF nanocomposites had excellent electrical conductivity, large specific surface area, abundant active sites, and excellent electrochemical properties. The sensitive electrochemical determination of chloramphenicol (CAP) was achieved by constructing an electrochemical sensor using the MXene/CNF composite. The response current values of the MXene/CNF sensor exhibited a good linear response with CAP concentration ranging from 0.03-25μM, with a low detection limit of 9 nM. In addition, the sensor demonstrated good repeatability and reproducibility with relative standard deviations of 2.94 and 3.29%, respectively. Remarkably, the developed sensor was successfully used for the real-time detection of CAP in milk and jasmine tea beverage and satisfactory rates were obtained.

Published
2024
Full Text
View/download PDF

44. Protic salt-based nitrogen-doped mesoporous carbon for simultaneous electrochemical detection of Cd(II) and Pb(II)

Author

Qin, Danfeng, Xu, Ruiyu, Shen, Hangjia, Mamat, Xamxikamar, Wang, Le, Gao, Shanshuang, Wang, Ying, Yalikun, Nuerbiya, Wågberg, Thomas, Zhang, Shiguo, Yuan, Qunhui, Li, Yongtao, Hu, Guangzhi, Qin, Danfeng, Xu, Ruiyu, Shen, Hangjia, Mamat, Xamxikamar, Wang, Le, Gao, Shanshuang, Wang, Ying, Yalikun, Nuerbiya, Wågberg, Thomas, Zhang, Shiguo, Yuan, Qunhui, Li, Yongtao, and Hu, Guangzhi

Abstract

Nitrogen-doped mesoporous carbon (NMC) derived from a single small-molecule protic salt (p-phenylenediamine bisulfate) is used for sensing toxic heavy metal ions. Using Nafion, bismuth and NMC to anchor the glassy carbon electrode surface, the fabricate electrode shows high sensitivity for detecting Cd(II) and Pb(II). The limits of detection (S/N = 3) are estimated to be 0.3 mu g L-1 for Cd(II) and 0.4 mu g L-1 for Pb(II), respectively, which are 10 and 25 times lower than the maximum acceptable content for drinking water recommended by the WHO. Furthermore, the sensor is successfully used to analyze Cd(II) and Pb(II) in tap-water with high anti-interference capability and good recovery.

Published
2017
Full Text
View/download PDF

45. Protic salt-based nitrogen-doped mesoporous carbon for simultaneous electrochemical detection of Cd(ii) and Pb(ii)

Author

Qin, Danfeng, primary, Xu, Ruiyu, additional, Shen, Hangjia, additional, Mamat, Xamxikamar, additional, Wang, Le, additional, Gao, Shanshuang, additional, Wang, Ying, additional, Yalikun, Nuerbiya, additional, Wagberg, Thomas, additional, Zhang, Shiguo, additional, yuan, Qunhui, additional, Li, Yongtao, additional, and Hu, Guangzhi, additional

Published
2017
Full Text
View/download PDF

46. In situ magnesiothermal synthesis of mesoporous MgO/OMC composite for sensitive detection of lead ions

Author

Shen, Hangjia, Qin, Danfeng, Li, Yuzhen, Li, Shouzhu, Yang, Chi, Yuan, Qunhui, Wågberg, Thomas, Hu, Guangzhi, Shen, Hangjia, Qin, Danfeng, Li, Yuzhen, Li, Shouzhu, Yang, Chi, Yuan, Qunhui, Wågberg, Thomas, and Hu, Guangzhi

Abstract

Mesoporous materials have exceptional properties for application owing to their ability to absorb and interact with guest species. A novel MgO/OMC composite with mesoporous structure was successfully synthesized via in situ magnesiothermic reduction. The structure was confirmed by N-2 sorption isotherms, X-ray diffraction pattern (XRD), transmission electron microscopy (TEM) and scanning electron microscopy (SEM). By using nafion and bismuth films as co-modifiers, the MgO/OMC composite material shows high sensitivity of 0.113 mu A.L mu g(-1) and lower background current for electrochemical determination of lead ion (Pb2+) by using anodic stripping voltammetry. The high specific surface area and good mass transfer on the proposed mesoporous material, as well as the outstanding adsorption abilities of MgO to metal ions and the excellent conductivity of the carbon skeleton contribute to the enhanced electrochemical response of Pb2+. As the stripping response of Bi/MgO/OMC-Nafion/PGE is highly linear (R-2 = 0.998) over a Pb2+ concentration range of 2 to 300 mg/L, it was successfully used to analyse Pb2+ in real tap-water samples with good recoveries.

Published
2016
Full Text
View/download PDF

50. An Electrochemical Sensor Based on Green ?-AlOOH-Carbonated Bacterial Cellulose Hybrids for Simultaneous Determination Trace Levels of Cd(II) and Pb(II) in Drinking Water

Author

Qin, Danfeng, Hu, Xun, Dong, Yemin, Mamat, Xamxikamar, Li, Yongtao, Wagberg, Thomas, and Hu, Guangzhi

Abstract

An eco-friendly g-AlOOH-carbonated bacterial cellulose (g-AlOOH-CBC) hybrids material was fabricated by simple pyrolysis and hydrothermal treatments. The obtained hybrids possess an intrinsic 3D nanofibrous structure decorated with chaff-like g-AlOOH particles. Owing to the good adsorption property and conductive, g-AlOOH-CBC hybrids were used to modified the glass carbon electrode (GCE) for simultaneous determination of Cd(II) and Pb(II) in aqueous samples by differential pulse anodic stripping voltammetry (DPASV) method. Various parameters affected Cd(II) and Pb(II) measurement were optimized. Under the optimal conditions, the limit of detection (S/N = 3) of the g-AlOOH-CBC modified electrode was evaluated to be 0.17 ug*L[?]1 for Cd(II) and 0.10 ug*L[?]1 for Pb(II) with the linear range of the calibration curves ranged 0.5-250 ug*L[?]1 for Cd(II) and Pb(II). Furthermore, the developed electrode was also successfully utilized for monitoring trace Cd(II) and Pb(II) in drinking water samples with satisfactory results.

Published
2018

Catalog

Books, media, physical & digital resources
See catalog results