Kecheng Zheng

Kecheng Zheng is currently a researcher at Ant Research working with Yujun Shen. My research focuses on computer vision and deep learning, particularly on multi-modal learning.

From Nov. 2018 to Jan. 2019, I was under the supervision of Wei Wei who is a research scientist in GOOGLE cloud.

From Sep. 2019 to May. 2020, I was an intern at JD AI Lab, working with Wu Liu.

From Jul. 2020 to Jan. 2021, I worked at Intelligent Multimedia Group (IMG) in MSRA as a research intern, under the supervision of Cuiling Lan.

From Mar. 2022 to Jul. 2022, I worked at Ant Research as a research intern working with Deli Zhao.

Email: zkechengzk@gmail.com  /  Google Scholar  /  Github

I'm interested in computer vision, representation learning and multi-modal learning.

We design an Unsupervised Pre-training framework for ReID based on the contrastive learning (CL) pipeline, dubbed UP-ReID.

We focus on handling well the CC-ReID problem under a more challenging setting, i.e., just from a single image, which enables high-efficiency and latency-free pedestrian identify for real-time surveillance applications.

We propose a simple yet effective Calibrated Feature Decomposition (CFD) module that focuses on improving the generalization capacity for person re-identification through a more judicious feature decomposition and reinforcement strategy.

We propose a novel Debiased Batch Normalization via Gaussian Process approach (GDNorm) for generalizable person re-identification, which models the feature statistic estimation from BN layers as a dynamically self-refining Gaussian process to alleviate the bias to unseen domain for improving the generalization.

We propose a novel modality-adaptive mixup and invariant decomposition (MID) approach for RGB-infrared person re-identification towards learning modality-invariant and discriminative representations.

We propose a network named Pose-Guided Feature Learning with Knowledge Distillation (PGFL-KD), where the pose information is exploited to regularize the learning of semantics aligned features but is discarded in testing.

We propose a Group-aware Label Transfer (GLT) algorithm, which enables the online interaction and mutual promotion of pseudo-label prediction and representation learning.

We propose a novel Spatial-Temporal Correlation and Topology Learning framework (CTL) to pursue discriminative and robust representation by modeling cross-scale spatial-temporal correlation.

We propose a Disentanglement-based Cross-Domain Feature Augmentation (DCDFA) strategy, where the augmented features characterize well the target and source domain data distributions while inheriting reliable identity labels.

We propose to estimate and exploit the credibility of the assigned pseudo-label of each sample to alleviate the influence of noisy labels.

We propose a novel memory enhanced embedding learning (MEEL) method for video-text retrieval.

We propose a novel hierarchical Gumbel attention network for text-based person search via Gumbel top-k re-parameterization algorithm.

We first illustrate that one of the main challenges in such a reasoning task is the presence of distracting features, which requires the learning algorithm to leverage counterevidence and to reject any of the false hypotheses in order to learn the true patterns.

We propose a stacked convolutional deep encoding network for video-text retrieval task, which considers to simultaneously encode long-range and short-range dependency in the videos and texts.

We propose a novel Layout-Aware Convolutional Neural Network (LA-Net) for accurate monocular depth estimation by simultaneously perceiving scene layout and local depth details.

CoDeF,

FastReID: a Pytorch Toolbox for General Instance Re-identification,

Carver: Benchmarking and Analyzing 3D-aware Image Synthesis with a Modularized Codebase,

Cones 2,

Cones,

Feel free to steal this website's source code.