Z
Zhanglin Peng
Researcher at SenseTime
Publications - 30
Citations - 647
Zhanglin Peng is an academic researcher from SenseTime. The author has contributed to research in topics: Normalization (statistics) & Artificial neural network. The author has an hindex of 12, co-authored 27 publications receiving 485 citations. Previous affiliations of Zhanglin Peng include Sun Yat-sen University.
Papers
More filters
Proceedings ArticleDOI
Deep Boosting: Layered Feature Mining for General Image Classification
TL;DR: A novel computational architecture for general image feature mining, which assembles the primitive filters into compositional features in a layer-wise manner, which is able to generate expressive image representations while inducing very discriminate functions for image classification.
Proceedings ArticleDOI
Active Domain Adaptation with Multi-level Contrastive Units for Semantic Segmentation
TL;DR: A novel Active Domain Adaptation scheme with Multi-level Contrastive Units (ADA-MCU) for semantic image segmentation with a simple pixel selection strategy followed with the construction of multi-level contrastive units is introduced to optimize the model for both domain adaptation and active supervised learning.
Journal ArticleDOI
Multi-Stage Spatio-Temporal Aggregation Transformer for Video Person Re-identification
TL;DR: Wang et al. as mentioned in this paper proposed a multi-stage Spatial-Temporal Aggregation Transformer (MSTAT) with two novel designed proxy embedding modules to address the above issue.
Proceedings ArticleDOI
Scheduling Large-scale Distributed Training via Reinforcement Learning
TL;DR: Policy schedular is proposed that determines the arguments of learning rate (lr) by reinforcement learning, significantly reducing costs to tune them and achieving superior performances on various tasks and benchmarks.
Patent
Normalization method, apparatus and device for deep neural network, and storage medium
TL;DR: In this article, the authors proposed a normalization method for deep neural networks, in which the normalization is carried out in the at least one dimension, and statistical information of each dimension is covered, so that good robustness is achieved for statistics in each dimension.