Meng Cao

3rd Year M.Sc. Student @ Peking University

Email: mengcao@pku.edu.cn

Github: https://github.com/rookiecm

Resume: CV

About me

I'm currently a third-year M.Sc. student in School of Electronic and Computer Engineering, Peking University (PKU), advised by Prof. Yuexian Zou. Before that in 2018, I received my B.Sc. degree in School of Hydropower and Information Engineering in Huazhong University of Science and Technology (HUST), where I am fortunate enough to work with Prof. Chaoshun Li.

My research interests lay in Computer Vision and Computer Graphics, especially in :

(1) Natural Language Video Localization.

(2) face and scene generation from 3D perspective.

(3) Scene Text Detection.

News

[Aug. 2021] One paper is accepted by EMNLP-2021. Coming soon.

[June. 2021] One paper is accepted by IEEE-TIP .

[Mar. 2021] One paper is accepted by CVPR 2021 .

[Mar. 2021] One paper is accepted by IEEE-TCSVT .

Preprints


Task-agnostic Temporally Consistent Facial Video Editing

PDF

Publications

All you need is a second look: Towards tighter arbitrary shape text detection

Meng Cao, Yuexian Zou.

IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2020)

PDF

GISCA: Gradient-Inductive Segmentation Network With Contextual Attention for Scene Text Detection

Meng Cao, Yuexian Zou, Dongming Yang, Chao Liu

IEEE Access

PDF

Experiences

Tencent AI Lab Research intern Jul 2019 - Aug 2020

A Unified Framework for Video Portrait Manipulation: We conduct a 3D-based framework for general video portrait manipulations including face swapping, face reenactment and other applications.

Alibaba Group Software engineering intern Mar 2018 - Aug 2018

Unbalanced sample training: For the actual corpus, the unbalanced sampling algorithm Borderline-SMOTE is used. The training model is adjusted using fastText toolbox to realize the search recall sorting algorithm.

Qihoo 360 Software engineering intern Nov 2017 - Jan 2018

Malicious domain name detection: Based on the characteristics of the DGA malicious domain name generation algorithm, LSTM is used to detect hidden malicious domain names.