Search

Your search keyword '"Song, Yanfei"' showing total 18 results
Start Over Author "Song, Yanfei" Publication Year Range Last 3 years
18 results on '"Song, Yanfei"'

1. SAM-Lightening: A Lightweight Segment Anything Model with Dilated Flash Attention to Achieve 30 times Acceleration

Author

Song, Yanfei, Pu, Bangzheng, Wang, Peng, Jiang, Hongxu, Dong, Dong, Cao, Yongxiang, and Shen, Yiqing

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Segment Anything Model (SAM) has garnered significant attention in segmentation tasks due to their zero-shot generalization ability. However, a broader application of SAMs to real-world practice has been restricted by their low inference speed and high computational memory demands, which mainly stem from the attention mechanism. Existing work concentrated on optimizing the encoder, yet has not adequately addressed the inefficiency of the attention mechanism itself, even when distilled to a smaller model, which thus leaves space for further improvement. In response, we introduce SAM-Lightening, a variant of SAM, that features a re-engineered attention mechanism, termed Dilated Flash Attention. It not only facilitates higher parallelism, enhancing processing efficiency but also retains compatibility with the existing FlashAttention. Correspondingly, we propose a progressive distillation to enable an efficient knowledge transfer from the vanilla SAM without costly training from scratch. Experiments on COCO and LVIS reveal that SAM-Lightening significantly outperforms the state-of-the-art methods in both run-time efficiency and segmentation accuracy. Specifically, it can achieve an inference speed of 7 milliseconds (ms) per image, for images of size 1024*1024 pixels, which is 30.1 times faster than the vanilla SAM and 2.1 times than the state-of-the-art. Moreover, it takes only 244MB memory, which is 3.5\% of the vanilla SAM. The code and weights are available at https://anonymous.4open.science/r/SAM-LIGHTENING-BC25/.

Published
2024

17. A novel colorimetric aptasensor for sensitive tetracycline detection based on the peroxidase-like activity of Fe3O4@Cu nanoparticles and "sandwich" oligonucleotide hybridization.

Author

Wang, Lijun, Zhou, Hong, Wu, Xue, Song, Yanfei, Huang, Yukun, Yang, Xiao, and Chen, Xianggui

Subjects
TETRACYCLINES, PEROXIDASE, TETRACYCLINE, NUCLEIC acid hybridization, NUCLEIC acid probes, NANOPARTICLES
Abstract

A novel colorimetric aptasensor has been developed for highly sensitive tetracycline (TC) detection based on the peroxidase-like activity of Fe3O4@Cu nanoparticles and "sandwich" oligonucleotide hybridization. The Fe3O4@Cu nanoparticles with high peroxidase-like activity were successfully synthesized under mild conditions. Then, a "sandwich" oligonucleotide hybridization probe (a short amino-modified complementary sequence of a portion of the TC aptamer (cDNA1), TC aptamers, and a long complementary to 5′ terminal TC aptamer sequence (cDNA2)) was created in 96-wells plates via DNA hybridization in the absence of TC from the detection system. The unique "sandwich" oligonucleotide hybridization probe adsorbed large numbers of Fe3O4@Cu nanozymes while further enhancing its peroxidase-like activity. Based on the 3,3′,5,5′-tetramethylbenzidine (TMB)-hydrogen peroxide (H2O2) reporting system, the blue color of the solution decreased linearly with the increase of TC concentration, ranging from 10−3 to 103 μg/L with an ultralow limit of detection (LOD) of 0.91 ng/L (2 pM). The proposed method was successfully applied to detect TC in spiked milk samples, with recoveries of 81.8 to 112%, demonstrating the excellent potential for highly sensitive TC detection in milk. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

18. Effects of implied social presence and interaction on attention in a virtual setting

Author

Song, YanFei

Subjects
visual search, implied social presence, competition, cooperation, attention
Abstract

Abstract: The social interactions that we engage in with those around us are crucial to our successful navigation of daily life, and the presence of others can largely influence our attentional allocation. However, less is known regarding social presence and interactions as they occur online; as the recent pandemic has led to a shift to online virtual social gatherings, it is even more imperative for us to understand the ways in which this can affect our attention. Here, we sought to understand whether implied social interaction and implied social presence can affect attention in a virtual visual search task. Implied social interaction was operationalized by telling participants they were cooperating and competing with a partner, while implied social presence was operationalized by the visual depiction of the other player on screen. To solidify our manipulation of implied social interaction, feedback was provided in the form of telling participants whether they were accurate or not, and their relative speed compared to the previous pair (cooperation) or compared to the opponent (competition). Our results indicate that, in line with prior work, participants who were told they were competing were significantly faster but less accurate than those who were told they were cooperating, suggesting that it is feasible to study implied social interactions using a virtual format. Although we did not find a significant interaction between this effect and our social presence manipulation, when comparing social presence to a baseline condition with no partner evoked, we found that participants prioritized accuracy when told they were cooperating and speed when told they were competing as compared to individual search. These findings suggest that the threshold to imply social presence may be lower than we thought, and that our cognitions regarding our social interactions can play a large role in affecting our attention.

Published
2021

Catalog

Books, media, physical & digital resources
See catalog results