Cofusing about incremental object detection.

[jinweiisgreat]

Thank you for your nice paper. But I don't know how to conduct incremental object detection.

• According to the open world setting. When the model labels sufficient unknown classes, the model will forward these unknown classes to humans to identify, right?
• So, between task2 and task3, we need a human to manually add the new classes identified by task2 to the model, right?
• In my understanding, the whole experiment process can be divided into two parts. The first stage is detecting new classes and the second stage is adding new classes to the model, right?
• So, there will be a process of human-computer interaction between two tasks, right?

Sorry, I am new to the field of incremental learning. Just confusing about this. Maybe I missed something in the article.

知道的也可以一起讨论， 就是我不清楚增量环节是怎么做的，task1是voc，这我们都知道；task2是识别多出来的类，那么在task3之前是不是要手动把task2的unknown给标注出来。这个手动标注是怎么做的？所以说整个实验是分成了4个task，在每一个task之间是暂停程序，人为的加标签进去？这么看起来挺繁琐的。

我不知道啊，我看实验部分也没说，是不是我看漏了？