12 Hits in 4.6 sec

AutoFlow: Learning a Better Training Set for Optical Flow [article]

Deqing Sun, Daniel Vlasic, Charles Herrmann, Varun Jampani, Michael Krainin, Huiwen Chang, Ramin Zabih, William T. Freeman, Ce Liu
2021 arXiv   pre-print
To automate the process, we present AutoFlow, a simple and effective method to render training data for optical flow that optimizes the performance of a model on a target dataset.  ...  Synthetic datasets play a critical role in pre-training CNN models for optical flow, but they are painstaking to generate and hard to adapt to new applications.  ...  Acknowledgements We would like to thank Shuyang Cheng, Ekin Dogus Cubuk, Alex Dosovitskiy, Rico Jonschkowski, David Kao, Ang Li, Aaron Sarna, Austin Stone, and Barret Zoph for helpful discussions and support  ... 
arXiv:2104.14544v1 fatcat:wd2r2aa6ozajfe25rawf5s5h5q

Perceiver IO: A General Architecture for Structured Inputs Outputs [article]

Andrew Jaegle and Sebastian Borgeaud and Jean-Baptiste Alayrac and Carl Doersch and Catalin Ionescu and David Ding and Skanda Koppula and Daniel Zoran and Andrew Brock and Evan Shelhamer and Olivier Hénaff and Matthew M. Botvinick and Andrew Zisserman and Oriol Vinyals and Joāo Carreira
2022 arXiv   pre-print
flow estimation with no explicit mechanisms for multiscale correspondence.  ...  A central goal of machine learning is the development of systems that can solve many problems in as many data domains as possible.  ...  ACKNOWLEDGMENTS We are grateful to Ankush Gupta and Adrià Recasens Continente for reviewing drafts of this paper and to Deqing Sun for sharing code and helpful advice on the optical flow experiments.  ... 
arXiv:2107.14795v3 fatcat:3fhsztbabjd3hcuft3qpoqibeq

Detail Preserving Residual Feature Pyramid Modules for Optical Flow [article]

Libo Long, Jochen Lang
2021 arXiv   pre-print
According to the particular modular structure of RFPM, we introduce a special transfer learning approach that can dramatically decrease the training time compared to a typical full optical flow training  ...  Feature pyramids and iterative refinement have recently led to great progress in optical flow estimation.  ...  In order to reduce the effort in retraining a complete optical flow method, we also propose a new transfer learning strategy which directly adds RFPM to a trained optical flow method and thereby significantly  ... 
arXiv:2107.10990v1 fatcat:2dnyjzmqazazjosc6n5oymy4c4

Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlation [article]

Jiankun Li, Peisen Wang, Pengfei Xiong, Tao Cai, Ziwei Yan, Lei Yang, Jiangyu Liu, Haoqiang Fan, Shuaicheng Liu
2022 arXiv   pre-print
In this paper, we propose a set of innovative designs to tackle the problem of practical stereo matching: 1) to better recover fine depth details, we design a hierarchical network with recurrent refinement  ...  ; 3) we introduce a new synthetic dataset with special attention to difficult cases for better generalizing to real-world scenes.  ...  Acknowledgement We thank all the reviewers for their valuable comments. This work was supported by the National Natural Science Foundation of China (NSFC) under grants No.61872067.  ... 
arXiv:2203.11483v1 fatcat:dlqfa53jmrfzhe3va3pq7u4n2a

GMFlow: Learning Optical Flow via Global Matching [article]

Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Dacheng Tao
Learning-based optical flow estimation has been dominated with the pipeline of cost volume with convolutions for flow regression, which is inherently limited to local correlations and thus is hard to address  ...  a self-attention layer for flow propagation.  ...  Introduction Since the pioneering learning-based work, FlowNet [10] , optical flow has been regressed with convolutions for a long time [15-17, 31, 36, 39, 46, 51] .  ... 
doi:10.48550/arxiv.2111.13680 fatcat:5o566gz4ybglzecybdn3yllh4i

Proceedings Of The 7Th Asian-Australasian Conference On Precision Agriculture [article]

W Nelson
2017 Zenodo  
Poster Development of a test bench for drip irrigation electronic valve Methods: A multistage pump (Model: DRL10-8; DOOCH, Korea), flow rate sensors (Model: E-MAG-I, AUTOFLOW, Korea), and pressure  ...  We show that training models using this approach outperform training a similar model from scratch, using the same data, for weed classification.  ...  Keynote Research and application of plant protection techniques and implements for rice production in China Zhiyan Zhou 1, 2, 3, 4 , Lian Hu 1,2,3 , Xiwen Luo* 1,2,3 , Ying Zang 1, 2, 3, 4 Rice  ... 
doi:10.5281/zenodo.1006669 fatcat:jb6sqa7ayngmlg7amwlpdkfmcy

Proceedings Of The 7Th Asian-Australasian Conference On Precision Agriculture [article]

W Nelson
2017 Zenodo  
Methods: A multistage pump (Model: DRL10-8; DOOCH, Korea), flow rate sensors (Model: E-MAG-I, AUTOFLOW, Korea), and pressure sensors (Model: A-10; WIKA, Germany) were used to construct the test bench.  ...  We show that training models using this approach outperform training a similar model from scratch, using the same data, for weed classification.  ...  This paper will use a simulation testing methodology on real hardware to compare two existing ML algorithms and a prototype implementation of a newly developed algorithm for their applicability to VRF  ... 
doi:10.5281/zenodo.1006670 fatcat:wupvtcswufg3bkpm3mdkrn5a4m

A Review of Field Measurement Studies on Thermal Comfort, Indoor Air Quality and Virus Risk

Christina Kakoulli, Alexis Kyriacou, Michalis P. Michaelides
2022 Atmosphere  
as the starting point for conducting future field measurement studies.  ...  The ongoing COVID-19 pandemic has also highlighted the need for real-life experimental data acquired through field measurement studies to help us understand and potentially control the impact of IEQ on  ...  without windows, and ways of signposting for better orientation within the campus.  ... 
doi:10.3390/atmos13020191 fatcat:tiupv2xc6zcfjjxpxx22gfd67y

History of Science and Technology

1991 Bulletin of Science, Technology & Society  
Programmers did not need to undergo lengthy training sessions or sit idle while they learned a "new" language.  ...  , the majority of the people involved in sp [structured programming]are interested only in learning how to program better.  ...  Appendix Two Original Research Below is pseudo code for a flight search program, first in the style of structured programming and then in the style of ad hoc coding of the 1960s.  ... 
doi:10.1177/027046769101100610 fatcat:xac6li3shraexpvdhi66lwu22u

Impact of Leucine and HMB on Acute Muscle Damage: Focus on Repair and Regeneration in vivo and in vitro

A Brown
The satellite cell is the driving factor for muscle regeneration, influenced by their internal and external environment.  ...  Results: Post eccentric exercise, at 48 h there was a significant (both P < 0.05) increase in muscle soreness and CK in young males.  ...  The exercise protocol was designed to apply to sporting settings, as multiple sports/training use repeated jumps, for instance: basketball, volleyball, football and plyometric training.  ... 
doi:10.24377/ljmu.t.00012613 fatcat:ch7jn4fnxzfx7npgq4wtirbl7e

The role of hypoxia on interstitial mitral valve cells

Kareem Salhiyyah, Adrian Chester
Mitral valve disease is a multifactorial process. The valve is a complex structure that contains an amalgam of extracellular proteins, cellular components, nerves and blood vessels.  ...  Hypoxia could play a role in initiating the structural changes in the valve that lead to dysfunction of the valve.  ...  He enriched my surgical training. He very kindly provided me with precious surgical specimens for this thesis.  ... 
doi:10.25560/41078 fatcat:zu7w52uh65fj7gx75xasn4sx3u

Toll-Like Receptor 2 and Partner Receptors in Alzheimers Disease [article]

Liu Shirong, Universität Des Saarlandes, Universität Des Saarlandes
Determine the optical density of each well immediately using a microplate reader; set to 450 nm and wavelength correction to 570 nm. Subtract readings at 570 nm from the readings at 450 nm. 14.  ...  The mean fluorescence intensity (mFI) of internalized FITC-labeled Aβ42 was immediately determined by BD FACSCanto II flow cytometry (Franklin Lakes, NJ) (the use of the FACS machine was based on training  ...  The secondary structure of DD loop is critical in the TIR-TIR platform formation, which is essential for the dimerization of TLRs (Gautam et al., 2006) .  ... 
doi:10.22028/d291-21508 fatcat:dvvnuudnznai3fqdnajkpsygfu