This work is the first to consider a new type of bias - task-induced bias in a causal perspective, which is derived from special settings of continual learning. We analyzed three continual scenarios in a causal framework, and found that the task-induced bias is particularly detrimental to CIL. Based on this framework, we designed a causal intervention operation and implemented it as a causal debias module by exploiting a powerful attention mechanism.

In this work, we research a new task: cross-domain object detection with missing classes in target domain, where there are only partial classes have images and annotations in the target domain.

In this paper, we investigated how to control the information diffusion over networks through the use of the irrational users.