Zifeng Wu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Zifeng Wu has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Andrew C. Stone
Department of Pharmacy, Wexner Medical Center, The Ohio State University, Columbus, USA
1 shared publication Anton van Den Hengel
1 shared publicationB Wang
Chinese Academy of Medical Sciences, Nanjing, China
1 shared publication- Bethany Young Holt
1 CAMI Health, Initiative for MPTs, Public Health Institute, Oakland, CA, United States
1 shared publication C Deal
NIH/DMID, Washington, DC, United States
1 shared publicationChunhua Shen
1 shared publication
Latest Publications
- Top Cited Papers
- 1 June 2019
- Published by Elsevier in Pattern Recognition
- Vol. 90, 119-133
Wider or Deeper: Revisiting the ResNet Model for Visual Recognition
- 1 June 2019
- Published by Elsevier in Pattern Recognition
- Vol. 90, 119-133
Top Cited PapersWider or Deeper: Revisiting the ResNet Model for Visual Recognition
- Top Cited Papers
- 15 August 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 28 (11) , 3174-3182
Estimating Depth From Monocular Images as Classification Using Deep Fully Convolutional Residual Networks
Depth estimation from single monocular images is a key component in scene understanding. Most existing algorithms formulate depth estimation as a regression problem due to the continuous property of depths. However, the depth value of input data can hardly be regressed exactly to the ground-truth value. In this article, we propose to formulate depth estimation as a pixel-wise classification task. Specifically, we first discretize the continuous ground-truth depths into several bins and label the bins according to their depth ranges. Then we solve the depth estimation problem as classification by training a fully convolutional deep residual network. Compared with estimating the exact depth of a single point, it is easier to estimate its depth range. More importantly, by performing depth classification instead of regression, we can easily obtain the confidence of a depth prediction in the form of probability distribution. With this confidence, we can apply an information gain loss to make use of the predictions that are close to ground-truth during training, as well as fully-connected conditional random fields (CRF) for post-processing to further improve the performance. We test our proposed method on both indoor and outdoor benchmark RGB-D datasets and achieve state-of-the-art performance.
- 15 August 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 28 (11) , 3174-3182
Top Cited PapersEstimating Depth From Monocular Images as Classification Using Deep Fully Convolutional Residual Networks
Depth estimation from single monocular images is a key component in scene understanding. Most existing algorithms formulate depth estimation as a regression problem due to the continuous property of depths. However, the depth value of input data can hardly be regressed exactly to the ground-truth value. In this article, we propose to formulate depth estimation as a pixel-wise classification task. Specifically, we first discretize the continuous ground-truth depths into several bins and label the bins according to their depth ranges. Then we solve the depth estimation problem as classification by training a fully convolutional deep residual network. Compared with estimating the exact depth of a single point, it is easier to estimate its depth range. More importantly, by performing depth classification instead of regression, we can easily obtain the confidence of a depth prediction in the form of probability distribution. With this confidence, we can apply an information gain loss to make use of the predictions that are close to ground-truth during training, as well as fully-connected conditional random fields (CRF) for post-processing to further improve the performance. We test our proposed method on both indoor and outdoor benchmark RGB-D datasets and achieve state-of-the-art performance.
- Open Access
- 13 July 2013
- Published by BMJ in Sexually Transmitted Infections
- Vol. 89 (Suppl 1) , A267.1-A267
P3.376 Multidisciplinary Approach to Contain HIV-1 and Other STIs in China: Multipurpose Prevention Technologies
Background In China, the HIV prevalence continues to rise despite increasing efforts to contain the epidemic, and other sexually transmitted infections (STIs) are also on the increase. Furthermore, antibiotic-resistant STI pathogens are on the increase and many STIs facilitate the acquisition and/or transmission of HIV-1. Multipurpose prevention technologies (MPTs), including drugs, devices and vaccines that simultaneously prevent unintended pregnancies, HIV and other STIs, would provide effective approaches to meet these public health challenges. Methods The Initiative for MPTs developed Target Product Profiles (TPPs) to define key product attributes appropriate for those regions of the world hardest hit by unintended pregnancy, HIV, and other STIs. The TPP considered priority combinations of indications, mechanisms of action, dosing and administration, and formulation and delivery. Current and historic data from China were reviewed to determine which product attributes in an MPT would best meet the needs in China. Results Sexual transmission of HIV-1 has become the major route for new infection in China, and other STIs have increased more than 40 times from 1985. The increase in STI rates is associated with population mobility, and the geographic distribution of STIs overlaps with that of HIV, suggesting the possibility of co-infection. The rate of unintended pregnancy is low in China, given the “one child” policy. After analysing the current and historic data, we propose that MPTs specifically addressing HIV and STIs would best meet the public health challenges in China, and the preferred dosing form would be long-acting implant or IUD-based delivery. Conclusion The continuing rise in prevalence and rates of STIs, including HIV poses serious public health challenges to China, where sustained economic development has resulted in social transformation and unprecedented population mobility. MPTs that specifically target HIV and other STIs could provide an effective approach to meeting these public health challenges in China.
- 13 July 2013
- Published by BMJ in Sexually Transmitted Infections
- Vol. 89 (Suppl 1) , A267.1-A267
Open AccessP3.376 Multidisciplinary Approach to Contain HIV-1 and Other STIs in China: Multipurpose Prevention Technologies
Background In China, the HIV prevalence continues to rise despite increasing efforts to contain the epidemic, and other sexually transmitted infections (STIs) are also on the increase. Furthermore, antibiotic-resistant STI pathogens are on the increase and many STIs facilitate the acquisition and/or transmission of HIV-1. Multipurpose prevention technologies (MPTs), including drugs, devices and vaccines that simultaneously prevent unintended pregnancies, HIV and other STIs, would provide effective approaches to meet these public health challenges. Methods The Initiative for MPTs developed Target Product Profiles (TPPs) to define key product attributes appropriate for those regions of the world hardest hit by unintended pregnancy, HIV, and other STIs. The TPP considered priority combinations of indications, mechanisms of action, dosing and administration, and formulation and delivery. Current and historic data from China were reviewed to determine which product attributes in an MPT would best meet the needs in China. Results Sexual transmission of HIV-1 has become the major route for new infection in China, and other STIs have increased more than 40 times from 1985. The increase in STI rates is associated with population mobility, and the geographic distribution of STIs overlaps with that of HIV, suggesting the possibility of co-infection. The rate of unintended pregnancy is low in China, given the “one child” policy. After analysing the current and historic data, we propose that MPTs specifically addressing HIV and STIs would best meet the public health challenges in China, and the preferred dosing form would be long-acting implant or IUD-based delivery. Conclusion The continuing rise in prevalence and rates of STIs, including HIV poses serious public health challenges to China, where sustained economic development has resulted in social transformation and unprecedented population mobility. MPTs that specifically target HIV and other STIs could provide an effective approach to meeting these public health challenges in China.