Combining ADDA with Deep CORAL: Unsupervised Domain Adaptation for Image Classification
We combine Adversarial Discriminative Domain Adaptation (ADDA) with Deep CORAL to allow ADDA better utilize the pretrained initialization. Vanilla ADDA diverses drastically from the initialization resulting much poorer results in early epochs comparing to the initialization. It requires sophisticated fine-tuning for ADDA to give satisfying results. With our novel modifications ADDA-CORAL can be trained extremely faster and yields better results.
Bohua Wan, Cong Mu, Ruzhang Zhao, Zhuoying Li (Ordered by alphabetic)
Dyadic Relational Graph Convolutional Networks for Skeleton-based Human Interaction Recognition
We apply Graph Convolutional Networks on skeleton-based human-human interaction recognitions. We designed a Relational Adjacency Matrix (RAM) to represent dynamic relational graphs on the two actor's skeletons.
Liping Zhu, Bohua Wan (Co-First), Chengyang Li, Gangyi Tian, Yi Hou, Kun Yuan
Pattern Recognition 115 (2021): 107920.