search

naoto0804 / cross-domain-detection

Cross-Domain Weakly-Supervised Object Detection through Progressive Domain Adaptation [Inoue+, CVPR2018].
https://naoto0804.github.io/cross_domain_detection/
422 stars 77 forks source link

Training with own dataset. #12

Closed monjurulkarim closed 5 years ago

monjurulkarim commented 5 years ago

Hi, Thanks for the amazing paper and this repository. I am trying to implement this repository with my own dataset. I have a different number of classes. Can you give me any idea how can I do that?

naoto0804 commented 5 years ago

Note that the classed to be detected in the target domain is a subset of a pre-trained object detector. In our paper, I used VOC2007-12 pre-trained model (20 classes).

To fully use the power of pre-trained detector, we didn't modify the network and train the model (in watercolor and comic setting, the dataset only contains 6 class). During training, if there's no instance for a class, the model will (hopefully) less likely predict it.